Difference between revisions of "Google Cloud Platform"

From Christoph's Personal Wiki
Jump to: navigation, search
(Cloud Storage)
(Command Line Interface (CLI))
 
(38 intermediate revisions by the same user not shown)
Line 113: Line 113:
 
*** Cloud Identity (organization domain that is not a Google domain/account)
 
*** Cloud Identity (organization domain that is not a Google domain/account)
 
** Service account
 
** Service account
*** Special type of Google account that belongs to your application, not an end user
+
*** Special type of Google account that belongs to your application, not an end-user
*** Does not user username/password; uses encryption keys
+
*** Does not use usernames/passwords; uses encryption keys
 
*** Identity for carrying out server-to-server interactions in a project (e.g., local server back application writing data to Cloud Storage)
 
*** Identity for carrying out server-to-server interactions in a project (e.g., local server back application writing data to Cloud Storage)
 
*** Identified by an email address:
 
*** Identified by an email address:
Line 122: Line 122:
 
**** User-created (custom)
 
**** User-created (custom)
 
**** Built-in (Compute Engine and App Engine default service accounts)
 
**** Built-in (Compute Engine and App Engine default service accounts)
**** Google APIs service accounts (runs internal Google processes on your behalf
+
**** Google APIs service accounts (runs internal Google processes on your behalf)
 
*** Application access
 
*** Application access
 
*** Used to authenticate from one service to another:
 
*** Used to authenticate from one service to another:
Line 222: Line 222:
  
 
* Custom Roles
 
* Custom Roles
** Can only be used at the project or organization levels (the cannot be used at the folder level)
+
** Can only be used at the project or organization levels (they cannot be used at the folder level)
 
** If you want to give custom permissions to a Compute Engine VM, use a service account
 
** If you want to give custom permissions to a Compute Engine VM, use a service account
  
Line 302: Line 302:
 
** Execute requests for any method and see responses in real time
 
** Execute requests for any method and see responses in real time
 
** Easily make authenticated and authorized API calls
 
** Easily make authenticated and authorized API calls
 
===Cloud Marketplace===
 
 
* Formerly called "Cloud Launcher"
 
* A solution marketplace containing pre-packaged, ready-to-deploy solutions
 
** Some offered by Google
 
** Others by third-party vendors
 
  
 
==Google Compute==
 
==Google Compute==
Line 341: Line 334:
 
* Storage for VMs
 
* Storage for VMs
 
** Persistent disks (either standard or SSD)
 
** Persistent disks (either standard or SSD)
** Any data save to scratch space (local SSD) will not be saved when the VM is terminated
+
** Any data saved to scratch space (local SSD) will not be saved when the VM is terminated
  
 
* Preemtible VM
 
* Preemtible VM
Line 352: Line 345:
  
 
; Compute Engine Metadata
 
; Compute Engine Metadata
* The project metadata URL is: <code><nowiki>http://metadata.google.internal/computeMetadata/v1/project/</nowiki></code>
+
 
* The instance metadata URL is: <code><nowiki>http://metadata.google.internal/computeMetadata/v1/instance/</nowiki></code>
+
* Project metadata - visible to all VMs in a project
 +
* VM metadata - private to the VM:
 +
** Instance hostname
 +
** External IP address
 +
** SSH keys
 +
** Project ID
 +
** Service account information and token
 +
* key-value pairs
 +
** Directories containing key-value pairs
 +
** Can return: specific value for a key, all keys in a directory, or a recursive list of keys
 +
 
 +
;Query metadata
 +
* Console, CloudShell, or API
 +
* From VM, using wget or curl
 +
** The project metadata URL is: <code><nowiki>http://metadata.google.internal/computeMetadata/v1/project/</nowiki></code>
 +
** The instance metadata URL is: <code><nowiki>http://metadata.google.internal/computeMetadata/v1/instance/</nowiki></code>
 +
* Using <code>gcloud</code>:
 +
** Project: <code>gcloud compute project-info describe</code>
 +
** Instance: <code>gcloud compute instances describe <INSTANCE></code>
  
 
Each URL returns a set of entries that can be appended to the URL. Project settings contain info (e.g., project ID). Instance settings contain info on disks, hostname, machine type, etc.
 
Each URL returns a set of entries that can be appended to the URL. Project settings contain info (e.g., project ID). Instance settings contain info on disks, hostname, machine type, etc.
 +
 +
;Detect when metadata has changed
 +
* Detect using <code>wait_for_change</code>parameter:
 +
$ curl "<nowiki>http://metadata/computeMetadata/v1/instance/tags?wait_for_change=true</nowiki>" -H "Metadata-Flavor: Google"
 +
:note: <code>timeout_sec</code> - returns if value changed after ''n'' seconds
 +
* Entity tags - HTTP ETag (used for webcache validation)
 +
** If metadata ETag differs from local version, then the latest value is returned immediately
 +
** Instances can use ETags to validate if they have the latest value
 +
 +
;Handle maintenance event
 +
* Retrieve scheduling options (availability policies)
 +
** <code>/scheduling</code> directory
 +
** <code>maintenance-event</code> attribute
 +
** Notifies when a maintenance even is ''about'' to occur
 +
** Value changes 60 seconds before a transparent maintenance event starts
 +
* Query periodically to trigger application code prior to a transparent maintenance event
 +
** Example: backup, logging, save state
 +
 +
; Custom metadata
  
 
One can also set one's own values so that one can use them in code on the VM.
 
One can also set one's own values so that one can use them in code on the VM.
Line 369: Line 399:
 
; Preemptible VMs
 
; Preemptible VMs
 
Preemptible VMs are highly affordable, short-lived compute instances suitable for batch jobs and fault-tolerant workloads. Preemptible VMs offer the same machine types and options as regular compute instances and last for up to 24 hours. If your applications are fault-tolerant and can withstand possible instance preemptions, then preemptible instances can reduce your Google Compute Engine costs significantly.
 
Preemptible VMs are highly affordable, short-lived compute instances suitable for batch jobs and fault-tolerant workloads. Preemptible VMs offer the same machine types and options as regular compute instances and last for up to 24 hours. If your applications are fault-tolerant and can withstand possible instance preemptions, then preemptible instances can reduce your Google Compute Engine costs significantly.
 +
 +
====Managed Instance Groups====
 +
 +
* Deploys identical instances, based on an instance template
 +
* Instance group can be resized
 +
* Manager ensures all instances are in "RUNNING" state
 +
* Typically used with autoscaler
 +
* Can be single zone or regional
 +
* In the UI, the instance template dialog looks and works exactly like creating an instance, except that it records the choices so it can repeat them.
 +
 +
====Autoscaling====
 +
 +
* Available as part of the Compute Engine API
 +
* Used to automatically scale number of instances in a ''managed instance group'' based on workload
 +
** Helps reduce costs by terminating instances when not required
 +
* Create one autoscaler per managed instance group
 +
* Autoscaler can be used with zone-based managed instance groups or regional managed instance groups
 +
* Autoscaler is fast (typically ~1 minute moving window)
 +
 +
; Autoscaling policies
 +
 +
Policies determine behaviour
 +
 +
Policy options:
 +
* Average CPU utilization
 +
** If average usage of total vCPU cores in instance group exceeds target, autoscaler adds more instances
 +
* HTTP load balancing serving capacity (defined in the backend service)
 +
** Maximum CPU utilization
 +
** Maximum requests per second/instance
 +
* Stackdriver standard and custom metrics
 +
 +
Multiple policies:
 +
* Autoscaler allows multiple policies (up to 5), but ''only'' for a single managed instance group.
 +
* Autoscaler handles mutiple policies by calculating recommended number of VMs for each policy and picking the policy that leaves the largest number of VMs in the group
 +
** Ensures enough VMs to handle application workloads and allows you to scale apps that have multiple possible bottlenecks.
 +
 +
* Example: Enable autoscaling for a managed instance group using CPU utilization:
 +
 +
$ gcloud compute instance-groups managed set-autoscaling example-managed-instance-group \
 +
    --max-num-replicas 20 \
 +
    --target-cpu-utilization 0.75 \
 +
    --cool-down-period 90
 +
 +
Note: the "cool down period" is the number of seconds the autoscaler should wait after a VM has started before the autoscaler starts collecting information from it.
 +
 +
; Configuring an autoscaler
 +
 +
# Create instance template (including startup, shutdown scripts) to automate tasks that must take place while instance is unattended:
 +
#* Software installation
 +
#* Software startup and shutdown
 +
#* Log archiving
 +
# Create managed instance group
 +
# Create autoscaler
 +
# Optionally, define multiple policies for autoscalers
  
 
===Google App Engine (GAE)===
 
===Google App Engine (GAE)===
Line 396: Line 480:
 
** Limits on third-party software
 
** Limits on third-party software
  
* Example App Engine ''Standard'' environment workflow (e.g., for you web app)
+
* Example App Engine ''Standard'' environment workflow (e.g., for your web app)
 
*# Develop and test the web app locally
 
*# Develop and test the web app locally
 
*# Use the SDK to deploy to App Engine (Project -> App Engine -> App Servers -> Application Instances)
 
*# Use the SDK to deploy to App Engine (Project -> App Engine -> App Servers -> Application Instances)
*# App Engine automatically scales and reliable serves your web application
+
*# App Engine automatically scales and reliably serves your web application
 
*#* App Engine can access a variety of services using dedicated APIs (e.g., NoSQL, Memcache, task queues, scheduled tasks, search, logs, etc.)
 
*#* App Engine can access a variety of services using dedicated APIs (e.g., NoSQL, Memcache, task queues, scheduled tasks, search, logs, etc.)
  
Line 588: Line 672:
  
 
There are several ways to bring data into Cloud Storage:
 
There are several ways to bring data into Cloud Storage:
* Online transfer: self-managed copies using CLI or drap-and-drop
+
* Online transfer: self-managed copies using CLI or drag-and-drop
* Storage Transfer Storage: scheduled, managed batch transfers
+
* [https://cloud.google.com/storage-transfer/docs/ Storage Transfer Service]: scheduled, managed batch transfers
 
* Transfer Appliance: rackable appliances to securely ship your data
 
* Transfer Appliance: rackable appliances to securely ship your data
  
Line 660: Line 744:
 
* Vertical scaling (read and write)
 
* Vertical scaling (read and write)
 
* Horizontal scaling (read)
 
* Horizontal scaling (read)
* Google security (network firewalls and encryption
+
* Google security (network firewalls and encryption)
  
 
; Use cases
 
; Use cases
Line 869: Line 953:
 
* VPCs belong to GCP projects, however, if you wish to establish connections between VPCs, you can use VPC peering.
 
* VPCs belong to GCP projects, however, if you wish to establish connections between VPCs, you can use VPC peering.
 
** If you want to use the full power of IAM to control ''who'' and ''what'' in one project can interact with a VPC in another project, use ''shared VPCs''.
 
** If you want to use the full power of IAM to control ''who'' and ''what'' in one project can interact with a VPC in another project, use ''shared VPCs''.
 +
 +
; Shared VPC
 +
 +
Share GCP VPC networks across projects in your Cloud organization using Shared VPC
 +
 +
Shared VPC allows:
 +
* Creation of a VPC network of RFC1918 IP spaces that associated projects can use
 +
* Project admins to create VMs in the shared VPC network spaces
 +
* Network and security admins to create VPNs and firewall rules, usable by the projects in the VPC network
 +
* Policies to be applied and enforced easily across a Cloud organization
 +
 +
; VPC Network Peering
 +
 +
VPC Network Peering provides the following advantages over using external IP addresses or VPNs to connect networks:
 +
* Network latency
 +
* Network security
 +
* Network cost
  
 
===Cloud Load Balancers===
 
===Cloud Load Balancers===
Line 902: Line 1,003:
 
| Supported on any port number
 
| Supported on any port number
 
| Used for the internal tiers of multi-tier applications
 
| Used for the internal tiers of multi-tier applications
 +
|-
 +
| Distributes HTTP(S) traffic among groups of instances based on:<br/>
 +
* Proximity to the user<br/>
 +
* Requested URL<br/>
 +
* Both
 +
| Distributes SSL traffic among groups of instances based on proximity to user.
 +
| Distributes TCP traffic among groups of instances based on proximity to user.
 +
|
 +
* Distributes traffic among a pool of instances within a region.
 +
* Can balance any kind of TCP/UDP traffic.
 +
| Distributes traffic from GCP VMs to a group of instances in the same region.
 
|}
 
|}
 
</div>
 
</div>
 
<br clear="all"/>
 
<br clear="all"/>
 +
 +
====Network Load Balancing====
 +
 +
; Target Pools
 +
 +
A Target Pool resource defines a group of instances that should receive incoming traffic from forwarding rules.
 +
 +
* Target pools can only be used with forwarding rules that handle: TCP/UDP traffic
 +
* You must create a target pool before you can use it with a forwarding rule.
 +
* Each project can have up to 50 target pools.
 +
* A target pool can have only one health check.
 +
** Network load balancing only supports <code>httpHealthChecks</code>.
 +
* Instances can be in different zones but must be in the same region; add to pool at creation or use Instance Groups.
 +
 +
====Internal Load Balancing====
 +
 +
Internal load balancing allows you to:
 +
* Load balance TCP/UDP traffic using a private frontend IP.
 +
* Load balance across instances in a region.
 +
* Configure health checking for your backends.
 +
* Get the benefits of a fully managed load balancing service that scales as you need to handle client traffic.
  
 
===Cloud DNS===
 
===Cloud DNS===
Line 910: Line 1,043:
 
Cloud DNS is highly available and scalable
 
Cloud DNS is highly available and scalable
  
 +
* '''100% SLA''' (only GCP service that offers this)
 
* Create managed zones, then add, edit, delete DNS records.
 
* Create managed zones, then add, edit, delete DNS records.
 
* Programmatically manage zones and records using RESTful API or CLI.
 
* Programmatically manage zones and records using RESTful API or CLI.
 +
* Lookup that translates symbolic names to IP addresses
 +
* High-performance DNS lookup for your users
 +
* Cost-effective for massive updates (millions of records)
 +
* Manage DNS records through API or Console
 +
* Request routed to the nearest location, reducing latency
 +
* Use cases:
 +
** DNS resolver for your company's users w/o managing your own servers
 +
** DNS propagation of company DNS records
 +
 +
; Cloud DNS managed zones
 +
* An abstraction that manages all DNS records for a single domain name
 +
* One project may have multiple managed zones
 +
* Must enable the Cloud DNS API in GCP Console first
 +
* <code>gcloud dns managed-zones ...</code>
 +
* Managed zones:
 +
** Permission controls at project level
 +
** Monitor propagation of changes to DNS name servers ([https://cloud.google.com/dns/docs/monitoring docs])
  
 
===Cloud Content Delivery Network (CDN)===
 
===Cloud Content Delivery Network (CDN)===
Line 917: Line 1,068:
 
* Uses Google's globally distributed edge caches to cache content close to yours.
 
* Uses Google's globally distributed edge caches to cache content close to yours.
 
* Or, you can use CDN Interconnect if you would prefer to use a different (non-GCP) CDN.
 
* Or, you can use CDN Interconnect if you would prefer to use a different (non-GCP) CDN.
 +
 +
===Cloud VPN===
 +
 +
* Securely connects your on-premises network to your GCP VPC network
 +
* Traffic travelling between the two networks is protected as it travels over the Internet:
 +
** Encrypted by one VPN gateway
 +
** Decrypted by the other VPN gateway
 +
* 99.9% SLA
 +
* Supports site-to-site VPN
 +
* Supports:
 +
** Static routes
 +
** Dynamic routes (Cloud Router)
 +
* Supports IKEv1 and IKEv2, using a shared secret
 +
* Uses Encapsulating Security Payload (ESP) in tunnel-mode with authentication
 +
 +
===Cloud Router===
 +
 +
* Provides BGP routing
 +
** Dynamically discovers and advertises routes
 +
* Supports graceful restart
 +
* Supports ECMP
 +
* Primary/backup tunnels for failover
 +
** MED
 +
** AS Path length
 +
** AS Prepend
 +
 +
===Cloud Interconnect===
 +
 +
* Enterprise-grade connection to GCP
 +
* Connect through a service provider's network
 +
* Provides dedicated bandwidth (50Mbps - 10Gbps)
 +
* Provides access to private (e.g., RFC1918) network addresses
 +
* Enables easy hybrid Cloud deployment
 +
* Does not require the use of and management of hardware VPN devices
 +
 +
===External Peering===
 +
 +
; Direct Peering
 +
* Connect to GCP through Google POPs
 +
* Provides ''N'' x 10G transport circuits for private Cloud traffic
 +
* BGP direct connect between your network and Google's network at Edge Network locations
 +
* Autonomous System numbers (AS) are exchanged via IXPs and some private facilities
 +
* Technical, commercial, and legal requirements
 +
 +
==Resource Manager==
 +
 +
: [https://cloud.google.com/resource-manager/ Official website]
 +
 +
Cloud Resource Manager allows you to hierarchically manage resources by project, folder, and organization.
 +
 +
; Billing and Resource Monitoring
 +
 +
* Organization contains all billing accounts
 +
* Project is associated with one billing account
 +
* Project accumulates consumption of all resources
 +
* A resource belongs to one, and only one, project
 +
* Resource consumption is measured on:
 +
** Rate of use/time
 +
** Number of items
 +
** Feature use
 +
 +
; Resource hierarchy
 +
 +
* Global
 +
** Images
 +
** Snapshots
 +
** Networks
 +
* Regional
 +
** External IP addresses
 +
* Zonal
 +
** Instances
 +
** Disks
 +
 +
; Project quotas
 +
 +
All resources are subject to project quotas (or limits)
 +
 +
* Quotas typically fall into one of three categories:
 +
*# How many resources you can create per project;
 +
*# How quickly you can make API requests in a project (rate limits); and
 +
*# Some quotas are per region
 +
* Quota examples:
 +
** 5 networks per project
 +
** 300 admin requests per minute (e.g., Cloud Spanner)
 +
** 24 CPUs region/project
 +
* Most quotas can be increased through a self-service form or a support ticket
 +
** IAM & admin -> Quotas
 +
 +
; Labels
 +
 +
* A utility for organizing GCP resources
 +
** Labels are key-value pairs
 +
** Attached to resources (e.g., VMs, disk, snapshots, images)
 +
** Can be created/applied via the Console, <code>gcloud</code>, or API
 +
* Example uses of labels:
 +
** Search and list all resources (inventory)
 +
** Filter resources (e.g., separate production from test)
 +
** Labels used in scripts
 +
* Label specification
 +
**  A label is a key-value pair
 +
** Label keys and non-empty label values can contain lowercase letters, digits, and hyphens; must start with a letter; and must end with a letter or digit. The regular expression is: <code>[a-z]([-a-z0-9]*[a-z0-9])</code>
 +
** The maximum length of label keys and values is 63 characters.
 +
** There can be a maximum of 64 labels per resource.
  
 
==Stackdriver==
 
==Stackdriver==
  
 
: [https://cloud.google.com/stackdriver/ Official website]
 
: [https://cloud.google.com/stackdriver/ Official website]
 +
 +
Overview:
 +
* Integrated monitoring, logging, diagnostics
 +
* Manages across platforms
 +
** GCP and AWS
 +
** Dynamic discovery of GCP with smart defaults
 +
** Open source agents and integrations
 +
* Access to powerful data and analytics tools
 +
* Collaborations with third-party software
  
 
Stackdriver provides services for:
 
Stackdriver provides services for:
Line 939: Line 1,202:
 
** Latency reporting and sampling
 
** Latency reporting and sampling
 
** Per-URL latency and statistics
 
** Per-URL latency and statistics
 +
 +
; Monitoring
 +
 +
* Install monitoring agent:
 +
$ curl -O <nowiki>https://repo.stackdriver.com/stack-install.sh</nowiki>
 +
$ sudo bash stack-install.sh --write-gcm
 +
 +
<pre>
 +
# To install the Stackdriver monitoring agent:
 +
$ curl -sSO <nowiki>https://dl.google.com/cloudagents/install-monitoring-agent.sh</nowiki>
 +
$ sudo bash install-monitoring-agent.sh
 +
 +
# To install the Stackdriver logging agent:
 +
$ curl -sSO <nowiki>https://dl.google.com/cloudagents/install-logging-agent.sh</nowiki>
 +
$ sudo bash install-logging-agent.sh
 +
</pre>
 +
 +
* Group metrics:
 +
** Aggregate metrics across a set of machines
 +
*** Dynamically defined
 +
*** Useful for high-change environments
 +
** Separate production from development
 +
** Filter Kubernetes Engine data by name and custom tags for cluster
 +
 +
* Custom metrics (e.g., in Python)
 +
 +
* Alerts &mdash; best practices
 +
** Alert on symptoms, not causes (e.g., monitor failing queries, not database down)
 +
** Use multiple channels
 +
*** Avoid a single point-of-failure in your alert strategy
 +
*** Multiple notification channels (e.g., email, SMS, Slack, etc.)
 +
** Customize alerts to audience needs
 +
*** Use descriptions to tell them what actions to take, what resources to examine.
 +
** Avoid noise
 +
*** Adjust monitoring so alerts are actionable, not dismissable.
 +
 +
; Logging
 +
 +
* Platform, system, and application logs
 +
** API to write to logs
 +
** 30-day retention + option to transfer to Cloud Storage
 +
* Log search/view/filter
 +
* Log-based metrics
 +
* Monitoring alerts can be set on log events
 +
* Data can be exported to BigQuery
 +
 +
* Viewing and exporting
 +
** Configure sinks for export
 +
** Stackdriver logs must have access to the resource
 +
*** Cloud Storage: storage and archiving
 +
*** BigQuery: analysis
 +
*** Cloud Pub/Sub: software integration
 +
** Export process
 +
*** As new log entries arrive, they are exported to sinks.
 +
*** Existing log entries, at the time the sink is created, are not exported.
 +
*** Cloud Storage entries are batched and sent out (~hourly).
 +
 +
* Exporting logs
 +
** Retain data longer by exporting to different storage classes of Cloud Storage
 +
** Search and analyze logs with BigQuery
 +
** Advanced visualization with Cloud Datalab
 +
** Stream logs to application or endpoint with Cloud Pub/Sub
 +
 +
; Error Reporting
 +
 +
Aggregate and display errors for running Cloud services
 +
* Error notifications
 +
* Error dashboards
 +
* Python, Java, JavaScript, Ruby, C#, PHP, and Go
 +
 +
Available for:
 +
* App Engine Standard
 +
* App Engine Flexible (beta)
 +
* Compute Engine (beta)
 +
* AWS EC2 (beta)
 +
* [Not available for] Kubernetes Engine
 +
 +
; Tracing
 +
 +
Tracing system:
 +
* Displays data in near real-time
 +
* Latency reporting
 +
* Per-URL latency sampling
 +
 +
Collects latency data on:
 +
* App Engine
 +
* Google HTTPS load balancers
 +
* Applications instrumented with the Stackdriver Trace SDKs
 +
 +
; Debugging
 +
 +
* Inspect an application without stopping it or slowing it down significantly
 +
* Can be used with App Engine Standard or Flexible, Compute Engine, or Kubernetes Engine
 +
* Python, Java, or Go
 +
* Debug snapshots:
 +
** Capture call stack and local variables of a running application
 +
* Debug log-points:
 +
** Inject logging into a service without stopping it
  
 
==Big Data==
 
==Big Data==
Line 1,146: Line 1,507:
  
 
Fully featured Git repositories hosted on GCP.
 
Fully featured Git repositories hosted on GCP.
 +
 +
* Private Git ''remote'' on GCP
 +
** IAM roles: owner, editor, viewer
 +
** Stackdriver integration
 +
** Console source code browser
 +
* Create repo
 +
$ gcloud init
 +
$ gcloud source repos create <REPO_NAME>
 +
* Can be a mirror of a hosted GitHub or BitBucket repo
  
 
===Deployment Manager===
 
===Deployment Manager===
Line 1,160: Line 1,530:
 
* Besides YAML, you can also use Python or Jinja2 templates
 
* Besides YAML, you can also use Python or Jinja2 templates
 
* Deployment Manager is available at no additional charge to Cloud Platform customers.
 
* Deployment Manager is available at no additional charge to Cloud Platform customers.
 +
 +
; Creating a Deployment Configuration
 +
 +
* Creating a configuration
 +
** <code>*.yaml</code> file defines the basic configuration
 +
** Include <code>import</code> at the top of the YAML file to expand to full-featured templates written in Python or Jinja2
 +
** Program configuration is bidirectional and interactive: receives data like <code>machine-type</code> and returns data like <code>ip-address</code>
 +
* Use "preview" to validate configuration before using it:
 +
$ gcloud deployment-manager deployments update my-deploy --config *.yaml --preview
 +
 +
; Template details
 +
 +
* 10 MB limit (neither the original configuration or the expanded one)
 +
* Python or Jinja2 templates cannot make any network or system calls (they will automatically be rejected)
 +
* Templates can be nested
 +
** Isolate specific functions into meaningful files
 +
** Create reusable assets
 +
** Example: a separate template for firewall rules
 +
* Templates have properties
 +
* Templates can use environment variables
 +
* Supports the startup script and metadata capabilities
 +
* Deployments can be updated (uses GCP API)
 +
** Add resources: default policy is <code>acquire</code> or <code>create</code> as needed
 +
** Remove resources: default policy is to <code>delete</code> the resource
 +
 +
SEE: [https://cloud.google.com/deployment-manager/docs/step-by-step-guide/create-a-template Step-by-step guide].
 +
 +
===Cloud Marketplace===
 +
 +
: [https://cloud.google.com/marketplace/ Official website]
 +
 +
* Formerly called "Cloud Launcher"
 +
* A solution marketplace containing pre-packaged, ready-to-deploy solutions (deployed via Deployment Manager)
 +
** Some offered by Google
 +
** Others by third-party vendors
 +
* Separate fees:
 +
** license fees for software
 +
** image usage fees
 +
* Image usage fee vs. separate license is up to vendor
 +
* Google updates images, but not running instances
 +
* Cloud technology partners
  
 
==Command Line Interface (CLI)==
 
==Command Line Interface (CLI)==
Line 1,210: Line 1,621:
 
* Initialize gcloud:
 
* Initialize gcloud:
 
  $ gcloud init
 
  $ gcloud init
 +
 +
* Get a list of GCP authorized accounts:
 +
$ gcloud auth list
 +
 +
* Set the active account:
 +
$ gcloud config set account <account>
 +
 +
* Get ''current'' account:
 +
$ gcloud config get-value account
  
 
* Get current gcloud configuration:
 
* Get current gcloud configuration:
Line 1,465: Line 1,885:
 
  SEE: [[Docker]] for more details.
 
  SEE: [[Docker]] for more details.
  
* Configure Docker to use the gcloud command-line tool as a credential helper:
+
; Overview
 +
 
 +
* Docker container images
 +
* Public and private container storage
 +
* Fast, scalable retrieval and deployment
 +
* Billed for storage and egress, not per image
 +
* Works with open and 3rd party continuous delivery systems
 +
* IAM roles
 +
* ACLs for access control
 +
 
 +
; Details
 +
 
 +
* Configure (local) Docker to use the gcloud command-line tool as a credential helper:
 
  $ gcloud auth configure-docker
 
  $ gcloud auth configure-docker
  
Line 1,842: Line 2,274:
 
| GET /favicon.ico HTTP/1.0              |        55845 |
 
| GET /favicon.ico HTTP/1.0              |        55845 |
 
+----------------------------------------+--------------+
 
+----------------------------------------+--------------+
 +
</pre>
 +
 +
===Cloud VPN===
 +
 +
This section will show how to setup a VPN between two subnets in different regions.
 +
 +
; Create VPC networks
 +
 +
* VPC#1
 +
** Name: vpn-network-1
 +
** Subnet creation mode: Custom
 +
** Name: subnet-a
 +
** Region: us-east1
 +
** IP address range: 10.5.4.0/24
 +
* VPC#2
 +
** Name: vpn-network-2
 +
** Subnet creation mode: Custom
 +
** Name: subnet-b
 +
** Region: europe-west1
 +
** IP address range: 10.1.3.0/24
 +
 +
; Create test VMs
 +
 +
One in both regions (VM#1: us-east1; VM#2: europe-west1)
 +
 +
; Create firewall rules
 +
 +
VPC network -> Firewall rules
 +
 +
* Allow traffic to vpn-network-1
 +
** Name: allow-icmp-ssh-network-1
 +
** Network: vpn-network-1
 +
** Targets: All instances in the network
 +
** Source filter: IP ranges
 +
** Source IP ranges: 0.0.0.0/0
 +
** Protocols and ports: Specified protocols and ports
 +
*** tcp -> 22
 +
*** other -> icmp
 +
 +
* Allow traffic to vpn-network-2
 +
** Name: allow-icmp-ssh-network-2
 +
** Network: vpn-network-2
 +
** Targets: All instances in the network
 +
** Source filter: IP ranges
 +
** Source IP ranges: 0.0.0.0/0
 +
** Protocols and ports: Specified protocols and ports
 +
*** tcp -> 22
 +
*** other -> icmp
 +
 +
; Create and prepare the VPN gateways
 +
 +
Create two VPN gateways, one in each region. Create forwarding rules for Encapsulating Security Payload (ESP), UDP:500, and UDP:4500 for each gateway.
 +
 +
* Create vpn-1 gateway:
 +
<pre>
 +
$ gcloud compute target-vpn-gateways \
 +
    create vpn-1 \
 +
    --network vpn-network-1  \
 +
    --region us-east1
 +
</pre>
 +
 +
* Create vpn-2 gateway:
 +
<pre>
 +
$ gcloud compute target-vpn-gateways \
 +
    create vpn-2 \
 +
    --network vpn-network-2  \
 +
    --region europe-west1
 +
</pre>
 +
 +
* Reserve a static IP for each network:
 +
<pre>
 +
$ gcloud compute addresses create --region us-east1 vpn-1-static-ip
 +
$ gcloud compute addresses list
 +
$ export STATIC_IP_VPN_1=<IP address for vpn-1>
 +
 +
$ gcloud compute addresses create --region europe-west1 vpn-2-static-ip
 +
$ gcloud compute addresses list
 +
$ export STATIC_IP_VPN_2=<IP address for vpn-2>
 +
</pre>
 +
 +
; Create forwarding rules for both vpn gateways
 +
 +
The forwarding rules forward traffic arriving on the external IP to the VPN gateway. It connects them together. Create three forwarding rules for the protocols necessary for VPN.
 +
 +
* Create ESP forwarding rule for vpn-1:
 +
<pre>
 +
$ gcloud compute \
 +
    forwarding-rules create vpn-1-esp \
 +
    --region us-east1  \
 +
    --ip-protocol ESP  \
 +
    --address ${STATIC_IP_VPN_1} \
 +
    --target-vpn-gateway vpn-1
 +
</pre>
 +
 +
* Create ESP forwarding rule for vpn-2:
 +
<pre>
 +
$ gcloud compute \
 +
    forwarding-rules create vpn-2-esp \
 +
    --region europe-west1  \
 +
    --ip-protocol ESP  \
 +
    --address ${STATIC_IP_VPN_2} \
 +
    --target-vpn-gateway vpn-2
 +
</pre>
 +
 +
* Create UDP500 forwarding for vpn-1:
 +
<pre>
 +
$ gcloud compute \
 +
    forwarding-rules create vpn-1-udp500  \
 +
    --region us-east1 \
 +
    --ip-protocol UDP \
 +
    --ports 500 \
 +
    --address ${STATIC_IP_VPN_1} \
 +
    --target-vpn-gateway vpn-1
 +
</pre>
 +
 +
* Create UDP500 forwarding for vpn-2:
 +
<pre>
 +
$ gcloud compute \
 +
    forwarding-rules create vpn-2-udp500  \
 +
    --region europe-west1 \
 +
    --ip-protocol UDP \
 +
    --ports 500 \
 +
    --address ${STATIC_IP_VPN_2} \
 +
    --target-vpn-gateway vpn-2
 +
</pre>
 +
 +
* Create UDP4500 forwarding for vpn-1:
 +
<pre>
 +
$ gcloud compute \
 +
    forwarding-rules create vpn-1-udp4500  \
 +
    --region us-east1 \
 +
    --ip-protocol UDP --ports 4500 \
 +
    --address ${STATIC_IP_VPN_1} \
 +
    --target-vpn-gateway vpn-1
 +
</pre>
 +
 +
* Create UDP4500 forwarding for vpn-2:
 +
<pre>
 +
$ gcloud compute \
 +
    forwarding-rules create vpn-2-udp4500  \
 +
    --region europe-west1 \
 +
    --ip-protocol UDP --ports 4500 \
 +
    --address ${STATIC_IP_VPN_2} \
 +
    --target-vpn-gateway vpn-2
 +
</pre>
 +
 +
* List forwarding rules (created above):
 +
<pre>
 +
$ gcloud compute forwarding-rules list
 +
NAME          REGION        IP_ADDRESS      IP_PROTOCOL  TARGET
 +
vpn-2-esp      europe-west1  34.76.244.229  ESP          europe-west1/targetVpnGateways/vpn-2
 +
vpn-2-udp4500  europe-west1  34.76.244.229  UDP          europe-west1/targetVpnGateways/vpn-2
 +
vpn-2-udp500  europe-west1  34.76.244.229  UDP          europe-west1/targetVpnGateways/vpn-2
 +
vpn-1-esp      us-east1      35.237.123.102  ESP          us-east1/targetVpnGateways/vpn-1
 +
vpn-1-udp4500  us-east1      35.237.123.102  UDP          us-east1/targetVpnGateways/vpn-1
 +
vpn-1-udp500  us-east1      35.237.123.102  UDP          us-east1/targetVpnGateways/vpn-1
 +
</pre>
 +
 +
* List VPN gateways:
 +
<pre>
 +
$ gcloud compute target-vpn-gateways list
 +
NAME  NETWORK        REGION
 +
vpn-2  vpn-network-2  europe-west1
 +
vpn-1  vpn-network-1  us-east1
 +
</pre>
 +
 +
; Create tunnels
 +
 +
Create the tunnels between the VPN gateways. After the tunnels exist, create a static route to enable traffic to be forwarded into the tunnel. If this is successful, you can ping a local VM in one location on its internal IP from a VM in a different location.
 +
 +
* Create the tunnel for traffic from Network-1 to Network-2:
 +
<pre>
 +
$ gcloud compute \
 +
    vpn-tunnels create tunnel1to2  \
 +
    --peer-address ${STATIC_IP_VPN_2} \
 +
    --region us-east1 \
 +
    --ike-version 2 \
 +
    --shared-secret gcprocks \
 +
    --target-vpn-gateway vpn-1 \
 +
    --local-traffic-selector 0.0.0.0/0 \
 +
    --remote-traffic-selector 0.0.0.0/0
 +
</pre>
 +
 +
* Create the tunnel for traffic from Network-2 to Network-1:
 +
<pre>
 +
$ gcloud compute \
 +
    vpn-tunnels create tunnel2to1 \
 +
    --peer-address ${STATIC_IP_VPN_1} \
 +
    --region europe-west1 \
 +
    --ike-version 2 \
 +
    --shared-secret gcprocks \
 +
    --target-vpn-gateway vpn-2 \
 +
    --local-traffic-selector 0.0.0.0/0 \
 +
    --remote-traffic-selector 0.0.0.0/0
 +
</pre>
 +
 +
* List VPN tunnels (created above):
 +
<pre>
 +
$ gcloud compute vpn-tunnels list
 +
NAME        REGION        GATEWAY  PEER_ADDRESS
 +
tunnel2to1  europe-west1  vpn-2    35.237.123.102
 +
tunnel1to2  us-east1      vpn-1    34.76.244.229
 +
</pre>
 +
 +
; Create static routes
 +
 +
* Create a static route from Network-1 to Network-2:
 +
<pre>
 +
$ gcloud compute  \
 +
    routes create route1to2  \
 +
    --network vpn-network-1 \
 +
    --next-hop-vpn-tunnel tunnel1to2 \
 +
    --next-hop-vpn-tunnel-region us-east1 \
 +
    --destination-range 10.1.3.0/24
 +
</pre>
 +
 +
* Create a static route from Network-2 to Network-1:
 +
<pre>
 +
$ gcloud compute  \
 +
    routes create route2to1  \
 +
    --network vpn-network-2 \
 +
    --next-hop-vpn-tunnel tunnel2to1 \
 +
    --next-hop-vpn-tunnel-region europe-west1 \
 +
    --destination-range 10.5.4.0/24
 +
</pre>
 +
 +
* List static routes (created above):
 +
<pre>
 +
$ gcloud compute routes list
 +
NAME        NETWORK        DEST_RANGE    NEXT_HOP                            PRIORITY
 +
...
 +
route1to2  vpn-network-1  10.1.3.0/24    us-east1/vpnTunnels/tunnel1to2      1000
 +
route2to1  vpn-network-2  10.5.4.0/24    europe-west1/vpnTunnels/tunnel2to1  1000
 +
</pre>
 +
 +
; Verify VPN connectivity
 +
 +
SSH into VM1 and check that you can ping the ''internal'' IP of VM2, and ''vice versa''.
 +
 +
===SSH into instance/VM===
 +
 +
SEE: [https://cloud.google.com/sdk/gcloud/reference/compute/ssh gcloud compute ssh] for details.
 +
 +
* To SSH into "<code>example-instance</code>" in zone <code>us-west2-a</code>, run:
 +
$ gcloud compute ssh example-instance --zone=us-west2-a
 +
 +
* You can also run a command on the virtual machine. For example, to get a snapshot of the guest's process tree, run:
 +
$ gcloud compute ssh example-instance --zone=us-west2-a --command="ps -ejH"
 +
 +
* If you are using the Google Container-Optimized virtual machine image, you can SSH into one of your containers with:
 +
$ gcloud compute ssh example-instance --zone=us-west2-a --container=CONTAINER
 +
 +
* You can limit the allowed time to ssh. For example, to allow a key to be used through 2019:
 +
$ gcloud compute ssh example-instance --zone=us-west2-a --ssh-key-expiration="2020-01-01T00:00:00:00Z"
 +
 +
* Or alternatively, allow access for the next two minutes:
 +
$ gcloud compute ssh example-instance --zone=us-west2-a --ssh-key-expire-after=2m
 +
 +
===Get billing information===
 +
<pre>
 +
$ gcloud alpha billing accounts list
 +
ACCOUNT_ID            NAME              OPEN  MASTER_ACCOUNT_ID
 +
000000-000000-000000  Blue Env - Xtof  True  0A0A0A-0B0B0B-0C0C0C
 +
 +
$ gcloud alpha billing projects list --billing-account=000000-000000-000000
 +
PROJECT_ID    BILLING_ACCOUNT_ID    BILLING_ENABLED
 +
blue-123456    AAAAAA-BBBBBB-CCCCCC  True
 +
green-654321  DDDDDD-EEEEEE-FFFFFF  False
 
</pre>
 
</pre>
  
 
==GCP vs. AWS==
 
==GCP vs. AWS==
 
''Note: All of the following are as of February 2017.''
 
''Note: All of the following are as of February 2017.''
 +
 +
:[https://cloud.google.com/docs/compare/aws/ GCP vs. AWS]
 +
:[https://cloud.google.com/docs/compare/azure/ GCP vs. Azure]
  
 
* Compute
 
* Compute
Line 2,038: Line 2,741:
 
==See also==
 
==See also==
 
* [[:Category:AWS|AWS]]
 
* [[:Category:AWS|AWS]]
 +
* [[Kubernetes]]
  
 
==External links==
 
==External links==
 
* [https://cloud.google.com/ Official website]
 
* [https://cloud.google.com/ Official website]
 +
* [https://cloud.google.com/docs/compare/aws#service_comparisons GCP ''vs.'' AWS]
  
 
[[Category:Technical and Specialized Skills]]
 
[[Category:Technical and Specialized Skills]]

Latest revision as of 00:23, 27 July 2021

Google Cloud Platform (GCP) is a cloud computing service by Google that offers hosting on the same supporting infrastructure that Google uses internally for end-user products like Gmail, Google Search, Maps, and YouTube.

Overview

  • Google Cloud Platform (GCP) high-level view:
    • Compute
      • App Engine (PaaS)
      • Kubernetes Engine (Hybrid; see Kubernetes)
      • Compute Engine (IaaS)
      • Cloud Functions
    • Storage
      • BigTable
      • Cloud Storage
      • Cloud SQL
      • Cloud Spanner
      • Cloud Datastore
    • Networking
      • VPCs
      • Load Balancers
      • Cloud DNS
      • Cloud CDN
    • Stackdriver
      • Monitoring
      • Logging
    • Big Data
      • BigQuery
      • Pub/Sub
      • Dataflow
      • Dataproc
      • Datalab
    • Artificial Intelligence
      • Natural Language API
      • Vision API
      • Speech API
      • Translate API
      • Machine Learning
Examples
  • Google Compute Engine – IaaS service providing virtual machines similar to Amazon EC2.
  • Google App Engine – PaaS service for directly hosting applications similar to AWS Elastic Beanstalk.
  • BigTable – IaaS service providing map reduce services. Similar to Hadoop.
  • BigQuery – IaaS service providing Columnar database. Similar to Amazon Redshift.
  • Google Cloud Functions – FaaS service allowing functions to be triggered by events without developer resource management similar to Amazon Lambda or IBM OpenWhisk.
Cloud Regions and Zones
  • Region
    • A Region is specific geographical location where you can run your resources
    • It is a collection of zones
    • Regional resources are available to resources in any zone in the region
    • They are frequently expanding
  • Zone
    • Zones are isolated physical locations within a region
    • Zonal resources are only available in that zone
    • Machines in different zones have no single point of failure

An effective disaster recovery plan would have assets deployed across multiple zones, or even different regions.

Standards, regulations, and certifications
GCP Certifications

Main

Cloud Resource Hierarchy
  • Provides a hierarchy of ownership
    • Identity and Access Management (IAM)
  • Provides "attach" points and inheritance for access control and organization policies
  • Hierarchy overview
    • Organization (not applicable to individual accounts)
    • Projects
    • Resources
Projects
  • Core organizational component of GCP
  • Controls access to resources (who has access to what)
  • Projects are where you create, enable, and use all GCP services
    • Per project basis
    • Permissions
    • Billing
    • APIs
    • Etc.
  • Projects have three identifying attributes:
    1. Project Name (user-friendly name)
    2. Project ID (aka Application ID; must be unique across GCP)
    3. Project Number (used in various places for identifying resources that belong to specific projects. For example, service account access names)

Identity and Access Management (IAM)

  • Who can do what on which resource
    • Members (who) are granted permissions and roles (what) to GCP services (resource) using the principle of least privilege
  • IAM -> Policy -> Roles + Identities
  • See predefined roles
  • Members (the "who")
    • Can be either a person or a service account
    • People via:
      • Google account
      • Google group (e.g., dev.team@thecompany.com)
      • G Suite Domain
      • Cloud Identity (organization domain that is not a Google domain/account)
    • Service account
      • Special type of Google account that belongs to your application, not an end-user
      • Does not use usernames/passwords; uses encryption keys
      • Identity for carrying out server-to-server interactions in a project (e.g., local server back application writing data to Cloud Storage)
      • Identified by an email address:
        • <project_number>@developer.gserviceaccount.com
        • <project_id>@developer.gserviceaccount.com
      • Three types of service accounts:
        • User-created (custom)
        • Built-in (Compute Engine and App Engine default service accounts)
        • Google APIs service accounts (runs internal Google processes on your behalf)
      • Application access
      • Used to authenticate from one service to another:
        • Programs running within Compute Engine instances can automatically acquire access tokens with credentials
        • Tokens used to access and service API in your project and any other services that granted access to that service account
        • Convenient when not accessing user data
  • Roles (the "what")
    • A collection of permissions to give access to a given resource
    • Permissions are represented as: <service>.<resource>.<verb> (e.g., compute.instances.delete)
    • Permissions vs. roles
      • Users are not directly assigned permissions, but are assigned roles, which contain a collection of permissions:
      Role                 List of permissions

                           compute.instances.delete
                           compute.instances.get
compute.instanceAdmin ---> compute.instances.list
                           compute.instances.setMachineType
                           compute.instances.start
                           compute.instances.stop
Cloud IAM objects
  • Organization (created by Google Sales)
    • Organization Owners are established at creation (note: always have more than one organization owner, for security purposes).
    • Organization Owner assigns the Organization Administrator role from the G Suite Admin Console (Admin is a separate product).
    • Organization Administrators manage GCP from the Cloud Console.
  • Folders
    • Additional grouping mechanism and isolation boundaries between projects (e.g., different departments or teams).
    • Folders allow delegation of administration rights.
  • Projects
  • Members
  • Roles
  • Resources
  • Products
  • G Suite Super Admins (are the only Organization Owners)
    • Administers a Google-hosted domain
    • Creates users, groups
    • Controls user membership in groups
Resource manager roles
  • Organization
    • Admin: full control over all resources
    • Viewer: view access to all resources
  • Folder
    • Admin: full control over folders
    • Creator: browse hierarchy and create folders
    • Viewer: view folders and projects below a resource
  • Project
    • Creator: create new projects (automatic owner) and migrate new projects into organization
    • Deleter: delete projects
Google Cloud Directory Sync (GCDS)
  • Synchronizes G Suite accounts to match the user data in existing LDAP or MS Active Directory
    • Syncs groups and memberships, not content or settings
    • Supports sophisticated rules for custom mapping of users, groups, non-employee contacts, user profiles, aliases, and exceptions
  • One-way synchronization from LDAP to directory
    • Administer in LDAP, then periodically update to G Suite
  • Runs as a utility in your server environment
Cloud IAM best practices
  • Principle of least privileges
    • Always apply the minimal access level required.
  • Use groups
    • If group membership is secure, assign roles to groups and let the G Suite Admins handle membership.
    • Always maintain an alternate.
    • For high-risk areas, assign roles to individuals directly and forego the convenience of group assignment.
  • Control who can change policies and group memberships
  • Audit policy changes
    • Audit logs record project-level permission changes.
    • Additional levels are being added all the time.
Primitive vs. Predefined (aka curated) vs. Custom Roles
  • Primitive Roles
    • Historically available GCP roles before Cloud IAM was implemented
    • Applied at Project-level
    • Broad roles:
    1. Viewer: read only actions that preserve state (i.e., cannot make changes)
    2. Editor: same as above + can modify state (e.g., deploy applications, modify code, configure services)
    3. Owner: same as above + can manage access to project and all project resources (e.g., invite/remove members and delete projects) + can setup project billing
    4. Billing administrator: manage billing + add/remove administrators
  • A project can have multiple owners, editors, viewers, and billing administrators
  • When to choose Primitive Roles
    • When the GCP service does not provide a predefined role
    • When you only need broad permissions for a project
    • When you want to allow a member to modify permissions for a project
    • When you work in a small team where the team members do not need granular permissions
  • Predefined ("Curated") Roles
    • Provides much more granular access (e.g., prevent unwanted access to other resources)
    • Granted at the resource-level
    • Example: App Engine Admin (full access to only App Engine resources)
    • Multiple predefined roles can be given to individual users
  • Custom Roles
    • Can only be used at the project or organization levels (they cannot be used at the folder level)
    • If you want to give custom permissions to a Compute Engine VM, use a service account
IAM Policy
  • A collection of statements that define who has what type of access
  • A full list of roles granted to a member for a resource
  • IAM Policy hierarchy
    • Resource access is organized hierarchically, from the Organization down to the Resource(s)
    • Organization -> Project -> Resource(s) — parent/child format
    • Each child has exactly one parent
    • Children inherit parent roles
    • Parent policies overrule restrictive child policies
Cloud Identity-Aware Proxy (Cloud IAP)
  • Enforce access control policies for applications and resources:
    • Identity-based access control
    • Central authorization layer for applications access by HTTPS
  • Cloud IAM policy is applied after authentication

Interacting with GCP

  • There are four methods of interacting with GCP:
    1. Cloud console (web user interface)
    2. Cloud Shell and Cloud SDK (command-line interface)
    3. Cloud Console Mobile App (for Android or iOS)
    4. RESTful API (for custom applications)
Google Cloud SDK
  • Command line interface (CLI) tools for managing resources and applications on GCP
  • Includes:
    • gcloud — many common GCP tasks
    • gsutil — interact with Cloud Storage
    • bq — interact with data in BigQuery
  • Can also be installed locally as a Docker image or run from within Cloud Shell (via the UI)
  • install
Cloud Shell
  • Interactive web-based shell environment for GCP, accessed from a web console
  • Easy to manage resources without having to install the Google Cloud SDK locally.
  • Includes:
    • A temporary Compute Engine virtual machine/instance
    • CLI access to the instance from a web browser
    • 5 GB of persistent disk storage
    • Pre-installed Google Cloud SDK and other tools
    • Language support for: Python, Go, Node.js, PHP, Ruby, and Java
    • Web preview functionality (especially useful for App Engine)
    • Built-in authorization for access to GCP projects and resources
  • Limitations
    • 1 hour time out for inactivity
      • Machine will terminate/self-delete
      • $HOME directory contents will be preserved for a new session
    • Direct interactive use only
      • Not for running high computational/network workloads
      • If in violation of GCP terms of use, session can be terminated without notice
    • For long periods of inactivity, home disk may be recycled (with advance notice via email)
      • If you need longer inactive period, consider either locally installed SDK or use Cloud Storage for long-term storage
RESTful APIs
  • "Intended for software developers" ~ Google
  • Programmatic access to GCP resources
    • Typically uses JSON as an interchange format
    • Uses OAuth 2.0 for authentication and authorization
  • Enabled via the GCP Console
  • Most APIs have daily quotas, which can be increased upon request (to Google Support)
  • You can experiment via APIs Explorer
APIs Explorer
  • An interactive tool that lets you easily try Google APIs using a browser
  • With the APIs Explorer, you can:
    • Browse quickly through available APIs and versions
    • See methods available for each API and what parameters they support, along with inline documentation
    • Execute requests for any method and see responses in real time
    • Easily make authenticated and authorized API calls

Google Compute

There are multiple Compute options in GCP for hosting your applications, where "option" is the method of hosting:

  • Google Compute Engine (GCE)
  • Google Container Engine (deprecated)
  • Google Kubernetes Engine (GKE)
  • Google App Engine (GAE)
  • Google Cloud Functions

In the above list, each option is ordered from "highly customizable" (GCE) to "highly managed" (Google Cloud Functions).

See: Choosing the right compute option in GCP: a decision tree

Each option can take advantage of the rest of the GCP services. E.g.,

  • Storage
  • Networking
  • Big Data
  • Security

Google Compute Engine (GCE)

Google Compute Engine (GCE)
  • Infrastructure as a Service (IaaS)
  • Virtual Machines (VMs), aka "instances"
  • Per-second billing; sustained use discounts
  • High throughput to storage at no extra cost
  • Custom machine types: Only pay for the hardware you need/use
  • Storage for VMs
    • Persistent disks (either standard or SSD)
    • Any data saved to scratch space (local SSD) will not be saved when the VM is terminated
  • Preemtible VM
Connecting to a VM
$ gcloud compute --project "my-project-123456" ssh --zone "us-west1-b" "my-vm"

The above command will create SSH keys (stored in ~/.ssh by default). After that, you can use the private key to SSH into your VM:

$ ssh -i ~/.ssh/google_compute_engine username@x.x.x.x
Compute Engine Metadata
  • Project metadata - visible to all VMs in a project
  • VM metadata - private to the VM:
    • Instance hostname
    • External IP address
    • SSH keys
    • Project ID
    • Service account information and token
  • key-value pairs
    • Directories containing key-value pairs
    • Can return: specific value for a key, all keys in a directory, or a recursive list of keys
Query metadata
  • Console, CloudShell, or API
  • From VM, using wget or curl
    • The project metadata URL is: http://metadata.google.internal/computeMetadata/v1/project/
    • The instance metadata URL is: http://metadata.google.internal/computeMetadata/v1/instance/
  • Using gcloud:
    • Project: gcloud compute project-info describe
    • Instance: gcloud compute instances describe <INSTANCE>

Each URL returns a set of entries that can be appended to the URL. Project settings contain info (e.g., project ID). Instance settings contain info on disks, hostname, machine type, etc.

Detect when metadata has changed
  • Detect using wait_for_changeparameter:
$ curl "http://metadata/computeMetadata/v1/instance/tags?wait_for_change=true" -H "Metadata-Flavor: Google"
note: timeout_sec - returns if value changed after n seconds
  • Entity tags - HTTP ETag (used for webcache validation)
    • If metadata ETag differs from local version, then the latest value is returned immediately
    • Instances can use ETags to validate if they have the latest value
Handle maintenance event
  • Retrieve scheduling options (availability policies)
    • /scheduling directory
    • maintenance-event attribute
    • Notifies when a maintenance even is about to occur
    • Value changes 60 seconds before a transparent maintenance event starts
  • Query periodically to trigger application code prior to a transparent maintenance event
    • Example: backup, logging, save state
Custom metadata

One can also set one's own values so that one can use them in code on the VM.

$ curl -H "Metadata-Flavor: Google" "http://metadata.google.internal/computeMetadata/v1/instance/service-accounts/default/token"
{"access_token":"aa00...","expires_in":3599,"token_type":"Bearer"}

$ curl -H "Metadata-Flavor: Google" "http://metadata.google.internal/computeMetadata/v1/project/"
attributes/
numeric-project-id
project-id
Preemptible VMs

Preemptible VMs are highly affordable, short-lived compute instances suitable for batch jobs and fault-tolerant workloads. Preemptible VMs offer the same machine types and options as regular compute instances and last for up to 24 hours. If your applications are fault-tolerant and can withstand possible instance preemptions, then preemptible instances can reduce your Google Compute Engine costs significantly.

Managed Instance Groups

  • Deploys identical instances, based on an instance template
  • Instance group can be resized
  • Manager ensures all instances are in "RUNNING" state
  • Typically used with autoscaler
  • Can be single zone or regional
  • In the UI, the instance template dialog looks and works exactly like creating an instance, except that it records the choices so it can repeat them.

Autoscaling

  • Available as part of the Compute Engine API
  • Used to automatically scale number of instances in a managed instance group based on workload
    • Helps reduce costs by terminating instances when not required
  • Create one autoscaler per managed instance group
  • Autoscaler can be used with zone-based managed instance groups or regional managed instance groups
  • Autoscaler is fast (typically ~1 minute moving window)
Autoscaling policies

Policies determine behaviour

Policy options:

  • Average CPU utilization
    • If average usage of total vCPU cores in instance group exceeds target, autoscaler adds more instances
  • HTTP load balancing serving capacity (defined in the backend service)
    • Maximum CPU utilization
    • Maximum requests per second/instance
  • Stackdriver standard and custom metrics

Multiple policies:

  • Autoscaler allows multiple policies (up to 5), but only for a single managed instance group.
  • Autoscaler handles mutiple policies by calculating recommended number of VMs for each policy and picking the policy that leaves the largest number of VMs in the group
    • Ensures enough VMs to handle application workloads and allows you to scale apps that have multiple possible bottlenecks.
  • Example: Enable autoscaling for a managed instance group using CPU utilization:
$ gcloud compute instance-groups managed set-autoscaling example-managed-instance-group \
    --max-num-replicas 20 \
    --target-cpu-utilization 0.75 \
    --cool-down-period 90

Note: the "cool down period" is the number of seconds the autoscaler should wait after a VM has started before the autoscaler starts collecting information from it.

Configuring an autoscaler
  1. Create instance template (including startup, shutdown scripts) to automate tasks that must take place while instance is unattended:
    • Software installation
    • Software startup and shutdown
    • Log archiving
  2. Create managed instance group
  3. Create autoscaler
  4. Optionally, define multiple policies for autoscalers

Google App Engine (GAE)

  • Platform as a Service (PaaS)
  • Developers can focus on writing code, while GCP handles the rest
  • Build scalable web applications and mobile backends
  • Your code is run as a binary called a "runtime"
  • It is a managed service
    • You never touch the underlying infrastructure
    • Deployment, maintenance, and scalability are handled for you
    • Reduces operational overhead
  • There are two environments available in GAE:
    1. Standard:
      • Simpler to use
      • Finer grain autoscale
      • Free daily quota
      • Supports specific versions of Python, PHP, Go, and Java
    2. Flexible:
      • (natively) supports Java 8, Servlet 3.1, Jetty 8, Python 2.7/3.5, Node.js, Ruby, PHP, .NET core, and Go (plus other runtimes, if using a custom Docker image)
App Engine Standard Environment
  • Your Standard App Engine Environment applications run in a "sandbox" and have the following constraints:
    • No writing to local files (must write to a DB instead for persistent data)
    • All requests your application receives have a 60 second time out
    • Limits on third-party software
  • Example App Engine Standard environment workflow (e.g., for your web app)
    1. Develop and test the web app locally
    2. Use the SDK to deploy to App Engine (Project -> App Engine -> App Servers -> Application Instances)
    3. App Engine automatically scales and reliably serves your web application
      • App Engine can access a variety of services using dedicated APIs (e.g., NoSQL, Memcache, task queues, scheduled tasks, search, logs, etc.)
App Engine Flexible Environment
  • Build and deploy containerized apps with a click
  • No sandbox constraints
  • Can access App Engine resources
  • Your apps run inside Docker containers on (managed) Google Compute Engine VMs.
    • Their health is monitored and automatically healed
    • You choose the geographical region
    • Critical backward compatible updates to the VM's OS is automatically applied
Comparison of App Engine environments
Standard Environment Flexible Environment
Instance startup: Milliseconds Minutes
SSH access No Yes (not by default)
Write to local disk No Yes (scratch space; writes are ephemeral)
Support for 3rd-party binaries No Yes
Network access Via App Engine Services Yes
Pricing model After free daily quota, pay per instance class, with automatic shutdown Pay for resource allocation per hour; no automatic shutdown


Google Cloud Functions

  • Serverless environment for building and connecting cloud services
  • Event-driven
    • Function executes as a "trigger" in response to a cloud-based event (Cloud Storage, Cloud Pub/Sub, or in an HTTP call)
    • Simple, single-purpose functions
  • Write function code in either Node.js or Python 3
  • Example: A file is uploaded to Cloud Storage (event), function executes in response to event (trigger)
  • Easier and less expensive than provisioning a server to watch for events
  • Not available in all Regions. As of January 2019, available in:
    • us-central1 (Iowa, USA)
    • us-east1 (South Carolina, USA)
    • europe-west1 (Belgium)
    • asia-northeast1 (Tokyo, Japan)

Comparing Compute Options

Comparison of Google Compute options
Compute Engine Kubernetes Engine App Engine Flexible App Engine Standard Cloud Functions
Service model IaaS Hybrid PaaS PaaS Serverless
Use cases General computing workloads Container-based workloads Web and mobile apps;
container-based workloads
Web and mobile apps Ephemeral functions responding to events
Toward managed infrastructure <--------------------> Toward dynamic infrastructure


Storage

Every application needs to store data, where it is business data, media to be streamd, or sensor data from devices. Consider the "three Vs" (3Vs):

  1. Variety: how similar structured or variable the data is;
  2. Velocity: how fast the data arrives; and
  3. Volatility: how long the data retains value and, therefore, needs to be accessible.

Cloud Storage

Cloud Storage is binary, large-object storage

  • High performance, Internet-scale
  • Data encryption at rest
  • Data encryption in transit by default from Google to endpoint
  • Bucket attributes
    • Globally unique name
    • Storage class
    • Location (region or multi-region)
    • IAM policies or Access Control Lists (ACLs)
    • Objects are immutable (turn on versioning for updating a "file" and keeping a history of changes)
    • Object lifecycle management rules (e.g., delete objects older than x-number of days; keep only the 3 most recent versions of an object {if versioning as been enabled on the bucket})
Cloud Storage Classes
Multi-regional Regional Nearline Coldline
Intended for data that is: Most frequently accessed Accessed frequently within a region Accessed less than once a month Accessed less than once a year
Feature Geo-redundant Regional Backup Archived or DR
Availability SLA 99.95% 99.90% 99.00% 99.00%
Durability 99.999999999% (11 nines) 99.999999999% (11 nines) 99.999999999% (11 nines) 99.999999999% (11 nines)
Duration Hot data Hot data 30-day minimum 90-day minimum
Access APIs Consistent APIs
Access time Millisecond access
Use cases Content storage and delivery In-region analytics; transcoding Long-tail content; backups Archiving; disaster recovery
Storage price $$$$ $$$ $$ $
Retrieval price none none $ $$


There are several ways to bring data into Cloud Storage:

  • Online transfer: self-managed copies using CLI or drag-and-drop
  • Storage Transfer Service: scheduled, managed batch transfers
  • Transfer Appliance: rackable appliances to securely ship your data

Cloud Storage works with other GCP services:

  • Compute Engine: VM startup scripts, images, and general object storage
  • App Engine: object storage, logs, and Datastore backups
  • Cloud SQL: import/export tables
  • BigQuery: import/export tables
Signed URLs
  • "Valet key" access to buckets and objects via ticket:
    • Ticket is a cryptographically signed URL
    • Time-limited
    • Operations specified in ticket: HTTP GET, PUT, DELETE (not POST)
    • Program can decide when or whether to give out signed URL
    • Any user with URL can invoke permitted operations
  • Example using private account key and gsutil
$ gsutil signurl -d 10m /path/to/privatekey.p12 gs://bucket/object

Example of a signed URL:

http://mybucket.storage.googleapis.com/foobar.txt?\
GoogleAccessId=1234567890123@developer.gserviceaccount.com&\
Expires=1552087696&\
Signature=ATlz98...asASDF345ASDF%3D
  • Query parameters:
    • GoogleAccessID: email address of the service account
    • Expires: when the signature expires (in Unix epoch format)
    • Signature: a cryptographic hash of composite URL string
  • The string that is digitally signed must contain:
    • HTTP verb (GET)
    • Expiration
    • Canonical resource (/bucket/object)

Cloud BigTable

Cloud BigTable is a fully managed NoSQL, wide-column database service for terabyte applications.

  • Accessed using the HBase API
  • Native compatibility with big data, Hadoop ecosystems
  • Managed, scalable storage
  • Data encryption in-flight and at rest
  • Control access with IAM
  • BigTable drives major applications, such as Google Search, Google Analytics, and Gmail
  • Learns and adjusts to access patterns
  • BigTable scales UP well; Datastore scales DOWN well.
  • If you need any of the following, consider using BigTable:
    • Storing > 1TB structure data;
    • Very high volume of writes;
    • Read/write latency < 10 milliseconds and strong consistency; and/or
    • HBase API compatible.
BigTable access patterns
  • Application API
    Data can be read from and written to Cloud BigTable through a data service layer, like Managed VMs, the HBase REST Server, or a Java Server using the HBase client. Typically, this will be to serve data to applications, dashboards, and data services.
  • Streaming
    Data can be streamed in (written even-by-even) through a variety of popular stream processing frameworks, like Cloud Dataflow Streaming, Spark Streaming, and Storm.
  • Batch Processing
    Data can be read from and written to Cloud BigTable through batch processes, like Hadoop MapReduce, Dataflow, or Spark. Often, summarized or newly calculated data is written back to Cloud BigTable or to a downstream database.

Cloud SQL

Cloud SQL is a managed RDBMS.

  • Offers MySQL and PostgeSQLBeta databases as a service (DBaaS)
  • Automatic replication
  • Managed backups (automatic or scheduled)
  • Vertical scaling (read and write)
  • Horizontal scaling (read)
  • Google security (network firewalls and encryption)
Use cases
  • App Engine
    Cloud SQL can be used with App Engine, using standard drivers.
    You can configure a Cloud SQL instance to follow an App Engine application.
  • Compute Engine
    Compute Engine instances can be authorized to access Cloud SQL instances using an external IP address.
    Cloud SQL instance can be configured with a preferred zone.
  • External services
    Cloud SQL can be used with external applications and clients.
    Standard tools can be used to administer databases.
    External read replicas can be configured.

Cloud Spanner

  • A horizontally scalable RDBMS (can scale to larger database sizes than Cloud SQL)
  • Transactional consistency at global scale
  • Managed instances with high availability
  • SQL queries (ANSI 2011 with extensions)
  • Automatic replication
  • Sharding
  • Use cases include financial applications and inventory applications

Cloud Datastore

Cloud Datastore is a horizontally scalable NoSQL DB.

  • Designed for application backends (databases can span Compute Engine and App Engine)
  • Scales automatically
  • Handles sharding and replication
  • Supports transactions that affect multiple database rows (unlike Cloud BigTable)
  • Allows for SQL-like queries
  • Includes a free daily quota (for storage, reads, writes, deletes, and small operations)

Comparing storage options

Cloud Datastore BigTable Cloud Storage Cloud SQL Cloud Spanner BigQuery1
Type NoSQL
document
NoSQL
wide column
Blobstore Relational
SQL for OLTP2
Relational
SQL for OLTP2
Relational
SQL for OLAP3
Access metaphor Persistent Hashmap Key-values, HBase API Like files in a file system Relational database Globally scalable RDBMS Relational
Read filter objects on property scan rows Must copy to local disk SELECT rows transactional reads/writes SELECT rows
Write put object put row One file INSERT row transactional reads/writes Batch/stream
Update granularity Attribute Row An object (a "file") Field SQL, Schemas ACID transactions
Strong consistency
High availability
Field
Transactions Yes Single-row No Yes Yes No
Complex queries No No No Yes Yes Yes
Capacity Terabytes+ Petabytes+ Petabytes+ Terabytes Petabytes Petabytes+
Unit size 1 MB/entry ~10 MB/cell
~100 MB/row
5 TB/object Determined by DB engine 10,240 MiB/row 10 MB/row
Best for Getting started, App Engine apps "Flat" data, heavy read/write, events, analytical data Structured and unstructured binary or object data Web frameworks, existing apps Large-scale database apps (> ~2 TB) Interactive querying, offline analytics
Usage Structured data from App Engine apps No-ops, high throughput, scalable, flattened data Store blobs No-ops SQL database SQL, Schemas ACID transactions
Strong consistency
High availability
Interactive SQL* querying fully managed warehouse
Use cases Getting started, App Engine apps AdTech, financial, and IoT data Images, large media files, backups User credentials, customer orders Whenever high I/O, global consistency is needed Data warehousing

1Sits on the edge of storage and data processing
2Online Transaction Processing (OLTP)
3Online Analytical Processing (OLAP)


See the Google storage decision tree for a graphical version of the above table.

Cloud Spanner Relational DB Non-relational DB
Schema Yes Yes No
SQL Yes Yes No
Consistency Strong Strong Eventual
Availability High Failover High
Scalability Horizontal Vertical Horizontal
Replication Automatic Configurable Configurable


Networking

Virtual Private Cloud (VPC)

Official Google Cloud VPC Product Page
  • Each VPC network is contained in a GCP project.
  • You can provision GCP resources, connect them to each other, and isolate them from one another.
  • Google Cloud VPC networks are global; subnets are regional (and subnets can span the zones that make up the region).
  • You can have resources in different zones on the same subnet.
  • You can dynamically increase the size of a subnet in a custom network by expanding the range of IP addresses allocated to it (without any workload shutdown or downtime).
  • Forward traffic from one instance to another instance within the same network, even across subnets, without requiring external IP addresses.
  • Use your VPC route table to forward traffic within the network, even across subnets (and zones) without requiring an external IP address.
  • VPCs give you a global distributed firewall.
  • You can define firewall rules in terms of metadata tags on VMs (e.g., tag all of your web servers {VMs} with "web" and write a firewall rule stating that traffic on ports 80 and/or 443 is allowed into all VMs with the "web" tag, no matter what their IP address happens to be).
  • VPCs belong to GCP projects, however, if you wish to establish connections between VPCs, you can use VPC peering.
    • If you want to use the full power of IAM to control who and what in one project can interact with a VPC in another project, use shared VPCs.
Shared VPC

Share GCP VPC networks across projects in your Cloud organization using Shared VPC

Shared VPC allows:

  • Creation of a VPC network of RFC1918 IP spaces that associated projects can use
  • Project admins to create VMs in the shared VPC network spaces
  • Network and security admins to create VPNs and firewall rules, usable by the projects in the VPC network
  • Policies to be applied and enforced easily across a Cloud organization
VPC Network Peering

VPC Network Peering provides the following advantages over using external IP addresses or VPNs to connect networks:

  • Network latency
  • Network security
  • Network cost

Cloud Load Balancers

With global Cloud Load Balancing, your application presents a single front-end to the world.

  • Users get a single, global anycast IP address.
  • Traffic goes over the Google backbone from the closest point-of-presence to the user.
  • Backends are selected based on load.
  • Only healthy backends receive traffic.
  • No pre-warming is required.
Cloud Load Balancing Options
Global HTTP(S) Global SSL Proxy Global TCP Proxy Regional Regional (internal)
Layer 7 load balancing based on load Layer 4 load balancing of non-HTTPS SSL traffic based on load Layer 4 load balancing of non-SSL TCP traffic Load balancing of any traffic (TCP, UDP) Load balancing of traffic inside a VPC
Can route different URLs to different backends Supported on specific port numbers Supported on specific port numbers Supported on any port number Used for the internal tiers of multi-tier applications
Distributes HTTP(S) traffic among groups of instances based on:
  • Proximity to the user
  • Requested URL
  • Both
Distributes SSL traffic among groups of instances based on proximity to user. Distributes TCP traffic among groups of instances based on proximity to user.
  • Distributes traffic among a pool of instances within a region.
  • Can balance any kind of TCP/UDP traffic.
Distributes traffic from GCP VMs to a group of instances in the same region.


Network Load Balancing

Target Pools

A Target Pool resource defines a group of instances that should receive incoming traffic from forwarding rules.

  • Target pools can only be used with forwarding rules that handle: TCP/UDP traffic
  • You must create a target pool before you can use it with a forwarding rule.
  • Each project can have up to 50 target pools.
  • A target pool can have only one health check.
    • Network load balancing only supports httpHealthChecks.
  • Instances can be in different zones but must be in the same region; add to pool at creation or use Instance Groups.

Internal Load Balancing

Internal load balancing allows you to:

  • Load balance TCP/UDP traffic using a private frontend IP.
  • Load balance across instances in a region.
  • Configure health checking for your backends.
  • Get the benefits of a fully managed load balancing service that scales as you need to handle client traffic.

Cloud DNS

Cloud DNS is highly available and scalable

  • 100% SLA (only GCP service that offers this)
  • Create managed zones, then add, edit, delete DNS records.
  • Programmatically manage zones and records using RESTful API or CLI.
  • Lookup that translates symbolic names to IP addresses
  • High-performance DNS lookup for your users
  • Cost-effective for massive updates (millions of records)
  • Manage DNS records through API or Console
  • Request routed to the nearest location, reducing latency
  • Use cases:
    • DNS resolver for your company's users w/o managing your own servers
    • DNS propagation of company DNS records
Cloud DNS managed zones
  • An abstraction that manages all DNS records for a single domain name
  • One project may have multiple managed zones
  • Must enable the Cloud DNS API in GCP Console first
  • gcloud dns managed-zones ...
  • Managed zones:
    • Permission controls at project level
    • Monitor propagation of changes to DNS name servers (docs)

Cloud Content Delivery Network (CDN)

  • Uses Google's globally distributed edge caches to cache content close to yours.
  • Or, you can use CDN Interconnect if you would prefer to use a different (non-GCP) CDN.

Cloud VPN

  • Securely connects your on-premises network to your GCP VPC network
  • Traffic travelling between the two networks is protected as it travels over the Internet:
    • Encrypted by one VPN gateway
    • Decrypted by the other VPN gateway
  • 99.9% SLA
  • Supports site-to-site VPN
  • Supports:
    • Static routes
    • Dynamic routes (Cloud Router)
  • Supports IKEv1 and IKEv2, using a shared secret
  • Uses Encapsulating Security Payload (ESP) in tunnel-mode with authentication

Cloud Router

  • Provides BGP routing
    • Dynamically discovers and advertises routes
  • Supports graceful restart
  • Supports ECMP
  • Primary/backup tunnels for failover
    • MED
    • AS Path length
    • AS Prepend

Cloud Interconnect

  • Enterprise-grade connection to GCP
  • Connect through a service provider's network
  • Provides dedicated bandwidth (50Mbps - 10Gbps)
  • Provides access to private (e.g., RFC1918) network addresses
  • Enables easy hybrid Cloud deployment
  • Does not require the use of and management of hardware VPN devices

External Peering

Direct Peering
  • Connect to GCP through Google POPs
  • Provides N x 10G transport circuits for private Cloud traffic
  • BGP direct connect between your network and Google's network at Edge Network locations
  • Autonomous System numbers (AS) are exchanged via IXPs and some private facilities
  • Technical, commercial, and legal requirements

Resource Manager

Official website

Cloud Resource Manager allows you to hierarchically manage resources by project, folder, and organization.

Billing and Resource Monitoring
  • Organization contains all billing accounts
  • Project is associated with one billing account
  • Project accumulates consumption of all resources
  • A resource belongs to one, and only one, project
  • Resource consumption is measured on:
    • Rate of use/time
    • Number of items
    • Feature use
Resource hierarchy
  • Global
    • Images
    • Snapshots
    • Networks
  • Regional
    • External IP addresses
  • Zonal
    • Instances
    • Disks
Project quotas

All resources are subject to project quotas (or limits)

  • Quotas typically fall into one of three categories:
    1. How many resources you can create per project;
    2. How quickly you can make API requests in a project (rate limits); and
    3. Some quotas are per region
  • Quota examples:
    • 5 networks per project
    • 300 admin requests per minute (e.g., Cloud Spanner)
    • 24 CPUs region/project
  • Most quotas can be increased through a self-service form or a support ticket
    • IAM & admin -> Quotas
Labels
  • A utility for organizing GCP resources
    • Labels are key-value pairs
    • Attached to resources (e.g., VMs, disk, snapshots, images)
    • Can be created/applied via the Console, gcloud, or API
  • Example uses of labels:
    • Search and list all resources (inventory)
    • Filter resources (e.g., separate production from test)
    • Labels used in scripts
  • Label specification
    • A label is a key-value pair
    • Label keys and non-empty label values can contain lowercase letters, digits, and hyphens; must start with a letter; and must end with a letter or digit. The regular expression is: [a-z]([-a-z0-9]*[a-z0-9])
    • The maximum length of label keys and values is 63 characters.
    • There can be a maximum of 64 labels per resource.

Stackdriver

Official website

Overview:

  • Integrated monitoring, logging, diagnostics
  • Manages across platforms
    • GCP and AWS
    • Dynamic discovery of GCP with smart defaults
    • Open source agents and integrations
  • Access to powerful data and analytics tools
  • Collaborations with third-party software

Stackdriver provides services for:

  • Monitoring
    • Platform, system, and application metrics
    • Uptime/health checks
    • Dashboards and alerts
  • Logging
    • Platform, system, and application logs
    • Log search, view, filter, and export
    • Log-based metrics
    • Export logs to BigQuery, Cloud Storage, and Cloud Pub/Sub
  • Debugger
    • Debug applications
  • Error reporting
    • Analyzes and aggregates the errors in your Cloud apps and notifies you when new errors are detected
  • Trace
    • Latency reporting and sampling
    • Per-URL latency and statistics
Monitoring
  • Install monitoring agent:
$ curl -O https://repo.stackdriver.com/stack-install.sh
$ sudo bash stack-install.sh --write-gcm
# To install the Stackdriver monitoring agent:
$ curl -sSO https://dl.google.com/cloudagents/install-monitoring-agent.sh
$ sudo bash install-monitoring-agent.sh

# To install the Stackdriver logging agent:
$ curl -sSO https://dl.google.com/cloudagents/install-logging-agent.sh
$ sudo bash install-logging-agent.sh
  • Group metrics:
    • Aggregate metrics across a set of machines
      • Dynamically defined
      • Useful for high-change environments
    • Separate production from development
    • Filter Kubernetes Engine data by name and custom tags for cluster
  • Custom metrics (e.g., in Python)
  • Alerts — best practices
    • Alert on symptoms, not causes (e.g., monitor failing queries, not database down)
    • Use multiple channels
      • Avoid a single point-of-failure in your alert strategy
      • Multiple notification channels (e.g., email, SMS, Slack, etc.)
    • Customize alerts to audience needs
      • Use descriptions to tell them what actions to take, what resources to examine.
    • Avoid noise
      • Adjust monitoring so alerts are actionable, not dismissable.
Logging
  • Platform, system, and application logs
    • API to write to logs
    • 30-day retention + option to transfer to Cloud Storage
  • Log search/view/filter
  • Log-based metrics
  • Monitoring alerts can be set on log events
  • Data can be exported to BigQuery
  • Viewing and exporting
    • Configure sinks for export
    • Stackdriver logs must have access to the resource
      • Cloud Storage: storage and archiving
      • BigQuery: analysis
      • Cloud Pub/Sub: software integration
    • Export process
      • As new log entries arrive, they are exported to sinks.
      • Existing log entries, at the time the sink is created, are not exported.
      • Cloud Storage entries are batched and sent out (~hourly).
  • Exporting logs
    • Retain data longer by exporting to different storage classes of Cloud Storage
    • Search and analyze logs with BigQuery
    • Advanced visualization with Cloud Datalab
    • Stream logs to application or endpoint with Cloud Pub/Sub
Error Reporting

Aggregate and display errors for running Cloud services

  • Error notifications
  • Error dashboards
  • Python, Java, JavaScript, Ruby, C#, PHP, and Go

Available for:

  • App Engine Standard
  • App Engine Flexible (beta)
  • Compute Engine (beta)
  • AWS EC2 (beta)
  • [Not available for] Kubernetes Engine
Tracing

Tracing system:

  • Displays data in near real-time
  • Latency reporting
  • Per-URL latency sampling

Collects latency data on:

  • App Engine
  • Google HTTPS load balancers
  • Applications instrumented with the Stackdriver Trace SDKs
Debugging
  • Inspect an application without stopping it or slowing it down significantly
  • Can be used with App Engine Standard or Flexible, Compute Engine, or Kubernetes Engine
  • Python, Java, or Go
  • Debug snapshots:
    • Capture call stack and local variables of a running application
  • Debug log-points:
    • Inject logging into a service without stopping it

Big Data

In the very near future, every company will be a data company, as making the fastest and best use of data is a critical source of competitive advantage.

Google Cloud's big data services are fully managed and scalable.

BigQuery 
Analytics database; stream data at 100,000 rows per seconds
Cloud Pub/Sub 
Scalable and flexible enterprise messaging
Cloud Dataproc 
Managed Hadoop MapReduce, Spark, Pig, and Hive service
Cloud Dataflow 
Stream and batch processing; unified and simplified pipelines
Cloud Datalab 
Interactive data exploration

BigQuery

Official website

BigQuery is a fast, highly scalable, cost-effective, and fully managed Cloud data warehouse for analytics, with built-in machine learning.

  • Provides near real-time interactive analysis of massive datasets (hundreds of TBs) using SQL syntax (SQL 2011)
  • Instead of using a dynamic pipeline (like Cloud Dataflow), use BigQuery for data that needs to run more in the way of exploring a vast sea of data (and are able to do ad hoc SQL queries on that massive dataset)
  • No cluster maintenance required
  • Load data from Cloud Storage or Cloud Datastore or stream it into BigQuery at up to 100,000 rows per second
  • In addition to SQL queries, you can read/write data in BigQuery via Cloud Dataflow, Hadoop, and Spark
  • Compute and storage are separated with a terabit network in between
  • You only pay for storage and processing use
    • You pay for your data storage separately from queries
  • Automatic discount for long-term data storage
    • When the age of your data reach 90 days in BigQuery, Google will automatically drop the price of storage.
  • Free monthly quotas
  • 99.9% SLA

Google's infrastructure is global and so is BigQuery. BigQuery lets you specify the region where your data will be kept. For example, if you want to keep data in Europe, you do not have to setup a cluster in Europe. Simply specify "EU" as the location when you create your dataset. US and Asia location are also available.

Cloud Pub/Sub

Official website

Cloud Pub/Sub allows you to ingest event streams from anywhere, at any scale, for simple, reliable, real-time stream analytics.

  • Scalable, reliable messaging
  • "Pub" => Publishers; "Sub" => Subscribers
  • Supports many-to-many asynchronous messaging
  • Application components make push/pull subscriptions to topics
  • Includes support for offline consumers
  • Designed to provide at least once delivery at low latency (i.e., it is possible for some messages to be delivered more than one; write your code to handle such situations)
  • Building block for data ingestion in Dataflow, IoT, Marketing Analytics, etc.
  • It is the foundation for Dataflow streaming
  • Useful for push notifications for cloud-based apps
  • Connect apps across GCP (e.g., push/pull between Compute Engine and App Engine)

Cloud Dataproc

Official website

Cloud Dataproc is a managed, Cloud-native Apache Hadoop & Apache Spark service.

  • A fast, easy, managed way to run Hadoop and Spark/Hive/Pig on GCP
  • Create clusters in 90 seconds or less (on average)
  • Scale clusters up and down, even when jobs are running
  • Easily migrate on-premises Hadoop jobs to the Cloud
  • Use Spark SQL and Spark Machine Learning libraries (MLlib) to run classification algorithms
  • Save money with preemptible instances
  • The rate for pricing is based on the hour, but Dataproc is billed by the second (one minute minimum)

The MapReduce module means that one function (traditionally called the "Map" function) runs in parallel with a massive dataset to produce intermediate results. Another function (the "Reduce" function) builds a final result set, based on all those intermediate results.

Cloud Dataflow

Official website

Cloud Dataflow is a simplified stream and batch data processing service, with equal reliability and expressiveness.

Use Cloud Dataproc when you have a dataset of known size or when you want to manage your cluster size yourself. If your data is ingested in real-time or is of an unpredictable size or rate, use Cloud Dataflow.

  • Offers managed data pipelines
  • Useful for fraud detection, financial services, IoT analytics, healthcare, logistics, clickstream, Point-of-Sale (PoS) and segmentation analysis in retail
  • Extract/Transform/Load (ETL) pipelines to move, filter, enrich, and shape data
  • Data analysis: batch computation and continuous computation using streaming
  • Processes data using Compute Engine instances
    • Clusters are sized for you
    • Automated scaling; no instance provisioning required
  • Write code once and get batch and streaming (transform-based programming model)
  • Orchestration: create pipelines that coordinate services, including external services
  • Integrates with GCP services: Cloud Storage, Cloud Pub/Sub, BigQuery, and BigTable
    • Open source Python and Java SDKs

Cloud Datalab

Official website

An interactive tool for data exploration, analysis, visualization, and machine learning.

  • Interactive tool for large-scale data exploration, transformation, analysis, and visualization
  • Integrated and open source (built on Jupyter)
  • Only pay for the resources you use (no charge for using Datalab itself)
  • Analyze data in BigQuery, Compute Engine, and Cloud Storage using Python, SQL, and JavaScaript
  • Easily deploy models to BigQuery
  • Visualize your data with Google Charts and matplotlib

Cloud Machine Learning

The Google Cloud Machine Learning Platform provides modern machine learning services with pre-trained models and a platform to generate your own taillored models.

  • Open source tool to build and run neural network models
    • Wide platform support: CPU, GPU, or TPU; mobile, server, or Cloud
  • Fully managed machine learning service
    • Familiar notebook-based developer experience
    • Optimized for Google's infrastructure; integrates with BigQuery and Cloud Storage
  • Pre-trained machine learning models built by Google
    • Speech: stream results in real-time; detects 80 languages
    • Vision: Identify objects, landmarks, text, and content
    • Translate: Language translation, including detection
    • Natural language: structure and meaning of text
Why use the Cloud Machine Learning Platform?
  • For structure data
    • Classification and regression
    • Recommendation
    • Anomaly detection
  • For unstructured data
    • Image and video analytics
    • Text analytics

Cloud Vision API

Official website

Analyze images with a simple REST API.

  • Logo detection, label detection, etc.
  • With the Cloud Vision API, you can:
    • Gain insight from images
    • Detect inappropriate context
    • Analyze sentiment
    • Extract text

Cloud Natural Language API

Cloud Speech-to-Text official website
  • Can return text in real-time
  • Highly accurate, even in noisy environments
  • Access from any device
  • As of March 2019, it recognizes over 120 languages and variants
  • Uses ML models to reveal structure and meaning of text
    • It can do syntax analysis (breaking down sentences into tokens, identify nouns, verbs, adjectives, and other parts of speech and figure out the relationships among the words).
  • Extract information about items mentioned in text documents, news articles, and blog posts

Cloud Translation API

Official website

Dynamically translate between languages.

  • Translate arbitrary strings between thousands of language pairs
  • Programmatically detect a document's language
  • Support for dozens of languages

Cloud Video Intelligence API

Official website

Search and discover your media content with Cloud Video Intelligence.

  • Annotate the contents of videos
  • Detect scene changes
  • Flag inappropriate content
  • Support for a variety of video formats

Tools

Cloud Endpoints

Official website

Develop, deploy, and manage APIs on any Google Cloud backend.

  • Distributed API management
  • Export your API using a RESTful interface
  • Control access and validate calls with JSON Web Tokens and Google API keys
    • Identify web / mobile users with Auth0 and Firebase Authentication
  • Generate client libraries
Supported platforms
  • Runtime environment
    • App Engine Flexible Environment
    • Kubernetes Engine
    • Compute Engine
  • Clients
    • Android
    • iOS
    • Javascript
Apigee Edge
  • A platform for making APIs available to your customers and partners
  • Helps you secure and monetize APIs
  • Contains analytics, monetization, and a developer portal

Cloud Source Repositories

Official website

Fully featured Git repositories hosted on GCP.

  • Private Git remote on GCP
    • IAM roles: owner, editor, viewer
    • Stackdriver integration
    • Console source code browser
  • Create repo
$ gcloud init
$ gcloud source repos create <REPO_NAME>
  • Can be a mirror of a hosted GitHub or BitBucket repo

Deployment Manager

Official website

Create and manage Cloud resources with simple templates (written in YAML).

  • It is an infrastructure management tool
  • Provides repeatable deployments
  • It is declarative; not imperative
    a declarative approach allows the user to specify what the configuration should be and let the system figure out the steps to take;
    an imperative approach requires the user to define the steps to take to create and configure resources
  • Besides YAML, you can also use Python or Jinja2 templates
  • Deployment Manager is available at no additional charge to Cloud Platform customers.
Creating a Deployment Configuration
  • Creating a configuration
    • *.yaml file defines the basic configuration
    • Include import at the top of the YAML file to expand to full-featured templates written in Python or Jinja2
    • Program configuration is bidirectional and interactive: receives data like machine-type and returns data like ip-address
  • Use "preview" to validate configuration before using it:
$ gcloud deployment-manager deployments update my-deploy --config *.yaml --preview
Template details
  • 10 MB limit (neither the original configuration or the expanded one)
  • Python or Jinja2 templates cannot make any network or system calls (they will automatically be rejected)
  • Templates can be nested
    • Isolate specific functions into meaningful files
    • Create reusable assets
    • Example: a separate template for firewall rules
  • Templates have properties
  • Templates can use environment variables
  • Supports the startup script and metadata capabilities
  • Deployments can be updated (uses GCP API)
    • Add resources: default policy is acquire or create as needed
    • Remove resources: default policy is to delete the resource

SEE: Step-by-step guide.

Cloud Marketplace

Official website
  • Formerly called "Cloud Launcher"
  • A solution marketplace containing pre-packaged, ready-to-deploy solutions (deployed via Deployment Manager)
    • Some offered by Google
    • Others by third-party vendors
  • Separate fees:
    • license fees for software
    • image usage fees
  • Image usage fee vs. separate license is up to vendor
  • Google updates images, but not running instances
  • Cloud technology partners

Command Line Interface (CLI)

The Google Cloud SDK is a set of tools that you can use to manage resources and applications hosted on the Google Cloud Platform (GCP). These include the gcloud, gsutil, and bq command line tools. The gcloud command-line tool is downloaded along with the Cloud SDK.

Configuration and services

$ gcloud components list
┌─────────────────────────────────────────────────────────────────────────────────────────────────────────────┐
│                                                  Components                                                 │
├───────────────┬──────────────────────────────────────────────────────┬──────────────────────────┬───────────┤
│     Status    │                         Name                         │            ID            │    Size   │
├───────────────┼──────────────────────────────────────────────────────┼──────────────────────────┼───────────┤
│ Not Installed │ App Engine Go Extensions                             │ app-engine-go            │  56.6 MiB │
│ Not Installed │ Cloud Bigtable Command Line Tool                     │ cbt                      │   6.4 MiB │
│ Not Installed │ Cloud Bigtable Emulator                              │ bigtable                 │   5.6 MiB │
│ Not Installed │ Cloud Datalab Command Line Tool                      │ datalab                  │   < 1 MiB │
│ Not Installed │ Cloud Datastore Emulator                             │ cloud-datastore-emulator │  17.7 MiB │
│ Not Installed │ Cloud Datastore Emulator (Legacy)                    │ gcd-emulator             │  38.1 MiB │
│ Not Installed │ Cloud Firestore Emulator                             │ cloud-firestore-emulator │  27.5 MiB │
│ Not Installed │ Cloud Pub/Sub Emulator                               │ pubsub-emulator          │  33.4 MiB │
│ Not Installed │ Cloud SQL Proxy                                      │ cloud_sql_proxy          │   3.8 MiB │
│ Not Installed │ Emulator Reverse Proxy                               │ emulator-reverse-proxy   │  14.5 MiB │
│ Not Installed │ Google Cloud Build Local Builder                     │ cloud-build-local        │   6.0 MiB │
│ Not Installed │ Google Container Registry's Docker credential helper │ docker-credential-gcr    │   1.8 MiB │
│ Not Installed │ gcloud Alpha Commands                                │ alpha                    │   < 1 MiB │
│ Not Installed │ gcloud Beta Commands                                 │ beta                     │   < 1 MiB │
│ Not Installed │ gcloud app Java Extensions                           │ app-engine-java          │ 107.5 MiB │
│ Not Installed │ gcloud app PHP Extensions                            │ app-engine-php           │           │
│ Not Installed │ gcloud app Python Extensions                         │ app-engine-python        │   6.2 MiB │
│ Not Installed │ gcloud app Python Extensions (Extra Libraries)       │ app-engine-python-extras │  28.5 MiB │
│ Not Installed │ kubectl                                              │ kubectl                  │   < 1 MiB │
│ Installed     │ BigQuery Command Line Tool                           │ bq                       │   < 1 MiB │
│ Installed     │ Cloud SDK Core Libraries                             │ core                     │   9.1 MiB │
│ Installed     │ Cloud Storage Command Line Tool                      │ gsutil                   │   3.5 MiB │
└───────────────┴──────────────────────────────────────────────────────┴──────────────────────────┴───────────┘
  • To install or remove components at your current SDK version [228.0.0], run:
$ gcloud components install COMPONENT_ID
$ gcloud components remove COMPONENT_ID
  • To update your SDK installation to the latest version [228.0.0], run:
$ gcloud components update
  • Initialize gcloud:
$ gcloud init
  • Get a list of GCP authorized accounts:
$ gcloud auth list
  • Set the active account:
$ gcloud config set account <account>
  • Get current account:
$ gcloud config get-value account
  • Get current gcloud configuration:
$ gcloud config list
[compute]
region = us-west1
zone = us-west1-a
[core]
account = someone@somewhere.com
disable_usage_reporting = True
project = my-project-223521

Your active configuration is: [default]
  • Get a list of all configurations:
$ gcloud config configurations list
NAME     IS_ACTIVE  ACCOUNT                PROJECT            DEFAULT_ZONE  DEFAULT_REGION
default  True       someone@somewhere.com  my-project-223521  us-west1-a    us-west1
  • Get a list of all (enabled) services:
$ gcloud services list
NAME                                    TITLE
bigquery-json.googleapis.com            BigQuery API
cloudapis.googleapis.com                Google Cloud APIs
clouddebugger.googleapis.com            Stackdriver Debugger API
cloudtrace.googleapis.com               Stackdriver Trace API
compute.googleapis.com                  Compute Engine API
container.googleapis.com                Kubernetes Engine API
containerregistry.googleapis.com        Container Registry API
datastore.googleapis.com                Cloud Datastore API
dns.googleapis.com                      Google Cloud DNS API
logging.googleapis.com                  Stackdriver Logging API
monitoring.googleapis.com               Stackdriver Monitoring API
oslogin.googleapis.com                  Cloud OS Login API
pubsub.googleapis.com                   Cloud Pub/Sub API
servicemanagement.googleapis.com        Service Management API
serviceusage.googleapis.com             Service Usage API
sql-component.googleapis.com            Cloud SQL
stackdriver.googleapis.com              Stackdriver API
stackdriverprovisioning.googleapis.com  Stackdriver Provisioning Service
storage-api.googleapis.com              Google Cloud Storage JSON API
storage-component.googleapis.com        Google Cloud Storage
$ gcloud services list --enabled --sort-by="NAME"
$ gcloud services list --available --sort-by="NAME"
Creating a project
  • Get a list of billing accounts:
$ gcloud beta billing accounts list
ACCOUNT_ID            NAME                OPEN  MASTER_ACCOUNT_ID
000000-000000-000000  My Billing Account  True
  • Create a project:
$ gcloud projects create dev-project-01 --name="dev-project-01" \
    --labels=team=area51
  • Link the above project to a billing account:
$ gcloud beta billing projects link dev-project-01 \
    --billing-account=000000-000000-000000
  • Switch between projects:
$ gcloud config set project ${PROJECT_NAME}
  • Get project-wide metadata (including project quotas):
$ gcloud compute project-info describe  # current project
#~OR~ specific project:
$ gcloud compute project-info describe --project ${PROJECT_NAME}
Managing multiple SDK configurations

Note: When you install the SDK, it will setup a default configuration and ask you to assign a project to it (and a default region).

  • Create a new configuration, activate, and switch between configurations:
$ gcloud config configurations create dev
$ gcloud config configurations list
$ gcloud config list
$ gcloud config configurations activate default
$ gcloud config set project dev-project-01
$ gcloud config set account someone@somewhere.com

Compute Engine

Creating a VM/instance
  • Use default values:
$ gcloud compute instances create "dev-server" --zone us-west1-a
  • Use customised values:
$ gcloud compute instances create "dev-server" \
    --project=my-project-123456 \
    --zone=us-west1-a \
    --machine-type=f1-micro \
    --subnet=default \
    --network-tier=PREMIUM \
    --maintenance-policy=MIGRATE \
    --service-account=00000000000-compute@developer.gserviceaccount.com \
    --scopes=https://www.googleapis.com/auth/devstorage.read_only,\
             https://www.googleapis.com/auth/logging.write,\
             https://www.googleapis.com/auth/monitoring.write,\
             https://www.googleapis.com/auth/servicecontrol,\
             https://www.googleapis.com/auth/service.management.readonly,\
             https://www.googleapis.com/auth/trace.append \
    --image=centos-7-v20181210 \
    --image-project=centos-cloud \
    --boot-disk-size=10GB \
    --boot-disk-type=pd-standard \
    --boot-disk-device-name=dev-server
  • Another example of creating a VM:
$ gcloud config list
$ gcloud compute zones list | grep us-west
$ gcloud config set compute/zone us-west1-a
$ gcloud compute images list --filter="debian"
$ gcloud compute instances create "my-vm-2" \
    --machine-type "n1-standard-1" \
    --image-project "debian-cloud" \
    --image "debian-9-stretch-v20190213" \
    --subnet "default"
Connecting to a VM (via SSH)
  • Google-managed:
$ gcloud compute instances list
$ gcloud compute ssh xtof@dev-server
$ gcloud compute ssh xtof@dev-server --dry-run  # see the actual command
  • Using your own SSH key:
$ ssh-keygen -t rsa -f my-ssh-key -C xtof
$ echo "xtof:$(cat my-ssh-key.pub)" > gcp_keys.txt
$ gcloud compute instances add-metadata dev-server --metadata-from-file ssh-keys=gcp_keys.txt
Snapshots
$ gcloud compute snapshots list
$ gcloud compute disks list
$ gcloud compute disks snapshot dev-server
$ gcloud compute snapshots delete <snapshot_name>
Images
  • Show public and private images (from which we can create instances from):
$ gcloud compute images list
NAME                 PROJECT        FAMILY     DEPRECATED   STATUS
centos-6-v20181210   centos-cloud   centos-6                READY
centos-7-v20181210   centos-cloud   centos-7                READY
...
Setting firewall rules
  • Get a list of current firewall rules (project-wide):
$ gcloud compute firewall-rules list 
NAME                    NETWORK  DIRECTION  PRIORITY  ALLOW                         DENY  DISABLED
default-allow-icmp      default  INGRESS    65534     icmp                                False
default-allow-internal  default  INGRESS    65534     tcp:0-65535,udp:0-65535,icmp        False
default-allow-rdp       default  INGRESS    65534     tcp:3389                            False
default-allow-ssh       default  INGRESS    65534     tcp:22                              False
  • Set firewall rules (e.g., allow HTTP/HTTPS traffic):
$ gcloud compute firewall-rules create default-allow-http \
    --project=my-project-123456 \
    --direction=INGRESS --priority=1000 --network=default \
    --action=ALLOW --rules=tcp:80 --source-ranges=0.0.0.0/0 \
    --target-tags=http-server

$ gcloud compute firewall-rules create default-allow-https \
    --project=my-project-123456 \
    --direction=INGRESS --priority=1000 --network=default \
    --action=ALLOW --rules=tcp:443 --source-ranges=0.0.0.0/0 \
    --target-tags=https-server
  • List updated firewall rules:
$ gcloud compute firewall-rules list 
NAME                    NETWORK  DIRECTION  PRIORITY  ALLOW                         DENY  DISABLED
default-allow-http      default  INGRESS    1000      tcp:80                              False
default-allow-https     default  INGRESS    1000      tcp:443                             False
...
  • Create an instance using the above firewall rules (HTTP/HTTPS):
$ gcloud compute instances create "dev-server" --zone us-west1-a \
    --tags=http-server,https-server
Deleting a VM/instance
  • Get a list of instances:
$ gcloud compute instances list
NAME        ZONE        MACHINE_TYPE  PREEMPTIBLE  INTERNAL_IP  EXTERNAL_IP    STATUS
dev-server  us-west1-a  f1-micro                   10.138.0.2   35.230.26.217  RUNNING
  • Delete the above instance:
$ gcloud compute instances delete "dev-server" --zone "us-west1-a"

Kubernetes

SEE: The Kubernetes main article.
Managing a GKE cluster
  • Create a basic Kubernetes cluster:
$ gcloud container clusters create my-k8s-cluster --zone us-west1-a --num-nodes 2
  • Create a Kubernetes cluster (with more options defined):
$ gcloud beta container --project "gcp-k8s-123456" clusters create "xtof-gcp-k8s" \
   --zone "us-west1-a" \
   --username "admin" \
   --cluster-version "1.11.5-gke.5" \
   --machine-type "n1-standard-1" \
   --image-type "COS" \
   --disk-type "pd-standard" \
   --disk-size "100" \
   --scopes \
     "https://www.googleapis.com/auth/devstorage.read_only",
     "https://www.googleapis.com/auth/logging.write",
     "https://www.googleapis.com/auth/monitoring",
     "https://www.googleapis.com/auth/servicecontrol",
     "https://www.googleapis.com/auth/service.management.readonly",
     "https://www.googleapis.com/auth/trace.append" \
   --num-nodes "3" \
   --enable-stackdriver-kubernetes \
   --no-enable-ip-alias \
   --network "projects/gcp-k8s-123456/global/networks/default" \
   --subnetwork "projects/gcp-k8s-123456/regions/us-west1/subnetworks/default" \
   --addons HorizontalPodAutoscaling,HttpLoadBalancing,KubernetesDashboard,Istio \
   --istio-config auth=NONE \
   --enable-autoupgrade \
   --enable-autorepair
  • Get the Kubernetes credentials:
$ gcloud container clusters get-credentials xtof-gcp-k8s --zone us-west1-a --project gcp-k8s-123456
  • Resize a Kubernetes cluster:
$ gcloud container cluster resize --size=1 --zone=us-west1-a xtof-gcp-k8s
  • Delete the cluster:
$ gcloud container clusters delete --project "gcp-k8s-123456" "xtof-gcp-k8s" --zone "us-west1-a"

Google Container Registry (GCR)

SEE: The Container Registry Quick Start guide for details.
SEE: Docker for more details.
Overview
  • Docker container images
  • Public and private container storage
  • Fast, scalable retrieval and deployment
  • Billed for storage and egress, not per image
  • Works with open and 3rd party continuous delivery systems
  • IAM roles
  • ACLs for access control
Details
  • Configure (local) Docker to use the gcloud command-line tool as a credential helper:
$ gcloud auth configure-docker

Note: The above command will add the following settings to your (local) Docker config file (located at ${HOME}/.docker/config.json):

 {
  "credHelpers": {
    "gcr.io": "gcloud", 
    "us.gcr.io": "gcloud", 
    "eu.gcr.io": "gcloud", 
    "asia.gcr.io": "gcloud", 
    "staging-k8s.gcr.io": "gcloud", 
    "marketplace.gcr.io": "gcloud"
  }
}
  • Tag the image with a registry name:
$ docker tag ${IMAGE_NAME} gcr.io/${PROJECT_ID}/${IMAGE_NAME}:${IMAGE_TAG}
  • Push the image to Container Registry:
$ docker push gcr.io/${PROJECT_ID}/${IMAGE_NAME}:${IMAGE_TAG}
  • List images in the Container Registry:
$ gcloud container images list
#~OR~
$ gcloud container images list --repository=gcr.io/${PROJECT_ID}
#~OR~
$ gcloud container images list --repository=gcr.io/${PROJECT_ID} --filter "name:${IMAGE_NAME}"
  • Pull the image from Container Registry:
$ docker pull gcr.io/${PROJECT_ID}/${IMAGE_NAME}:${IMAGE_TAG}
  • Cleanup (delete):
$ gcloud container images delete gcr.io/${PROJECT_ID}/${IMAGE_NAME}:${IMAGE_TAG} --force-delete-tags

Deployment Manager

"Deployment Manager is an infrastructure deployment service that automates the creation and management of Google Cloud Platform resources for you. Write flexible template and configuration files and use them to create deployments that have a variety of Cloud Platform services, such as Google Cloud Storage, Google Compute Engine, and Google Cloud SQL, configured to work together". source

  • Example:
$ gcloud deployment-manager deployments create my-deployment --config my-deployment.yml
$ gcloud deployment-manager deployments update my-deployment --config my-deployment.yml
$ gcloud deployment-manager deployments describe my-deployment

Miscellaneous

$ gcloud config set project <project-name>
$ gcloud config set compute/zone us-west1
$ gcloud config unset compute/zone
$ gcloud iam service-accounts list \
    --filter='displayName:"Compute Engine default service account"' \
    --format='value(email)'
$ gcloud iam service-accounts list --format=json | \
    jq -r '.[] | select(.email | startswith("my-project@")) | .email'
my-project@gmy-project-123456.iam.gserviceaccount.com
$ gcloud compute networks subnets list
NAME     REGION                   NETWORK  RANGE
default  us-west2                 default  10.168.0.0/20
default  asia-northeast1          default  10.146.0.0/20
default  us-west1                 default  10.138.0.0/20
default  southamerica-east1       default  10.158.0.0/20
default  europe-west4             default  10.164.0.0/20
default  asia-east1               default  10.140.0.0/20
default  europe-north1            default  10.166.0.0/20
default  asia-southeast1          default  10.148.0.0/20
default  us-east4                 default  10.150.0.0/20
default  europe-west1             default  10.132.0.0/20
default  europe-west2             default  10.154.0.0/20
default  europe-west3             default  10.156.0.0/20
default  australia-southeast1     default  10.152.0.0/20
default  asia-south1              default  10.160.0.0/20
default  us-east1                 default  10.142.0.0/20
default  us-central1              default  10.128.0.0/20
default  asia-east2               default  10.170.0.0/20
default  northamerica-northeast1  default  10.162.0.0/20
$ gcloud projects create example-foo-bar-1 --name="Happy project" \
    --labels=type=happy
$ gcloud compute forwarding-rules list \
    --filter='name:"my-app-forwarding-rules"' \
    --format='value(IPAddress)'
x.x.x.x
$ gcloud pubsub topics publish myTopic --message '{"name":"bob"}'
$ gcloud functions logs read

Cloud Storage

Storage Classes
Storage Class Name for APIs and gsutil
Multi-Regional Storage multi_regional
Regional Storage regional
Nearline Storage nearline
Coldline Storage coldline

See: for details


  • Create a bucket:
$ PROJECT_NAME=my-project
$ REGION=us-west1
$ STORAGE_CLASS=regional
$ BUCKET_NAME=xtof-test

# Basic (using defaults):
$ gsutil mb gs://${BUCKET_NAME}

# Advanced (override defaults):
$ gsutil mb -p ${PROJECT_NAME} -c ${STORAGE_CLASS} -l ${REGION} gs://${BUCKET_NAME}
# Use Cloud Shell variables:
$ gsutil mb -l US ${DEVSHELL_PROJECT_ID} # <- creates a globally unique bucket name based off of your project ID
# Set the ACL of an object in your bucket:
$ gsutil acl ch -u allUsers:R gs://${DEVSHELL_PROJECT_ID}/foobar.png

Note: All buckets (and their objects) are private by default.

  • Upload an object to the above bucket:
$ gsutil cp Pictures/foobar.jpg gs://${BUCKET_NAME}
  • Move an object (file) from one bucket to another:
$ gsutil mv gs://${SOURCE_BUCKET} gs://${DESTINATION_BUCKET}
  • List the contents of a bucket:
$ gsutil ls gs://${BUCKET_NAME}     # basic info
$ gsutil ls -l gs://${BUCKET_NAME}  # extended info
Identity and Access Management
  • Get the IAM roles and rules for a given bucket (note: these are the default ones):
$ gsutil iam get gs://${BUCKET_NAME}
{
  "bindings": [
    {
      "members": [
        "projectEditor:my-project-123456", 
        "projectOwner:my-project-123456"
      ], 
      "role": "roles/storage.legacyBucketOwner"
    }, 
    {
      "members": [
        "projectViewer:my-project-123456"
      ], 
      "role": "roles/storage.legacyBucketReader"
    }
  ], 
  "etag": "CAE="
}
Lifecycle Management
  • Find all objects in a given bucket older than 2 days (i.e., when they were uploaded to the bucket or last modified) and convert them from "regional" to "nearline" storage class:
$ cat << EOF > lifecycle.json
{
  "lifecycle": {
    "rule": [
      {
        "action": {
          "type": "SetStorageClass",
          "storageClass": "NEARLINE"
        },
        "condition": {
          "age": 2,
          "matchesStorageClass": [
            "REGIONAL"
          ]
        }
      }
    ]
  }
}
EOF

$ gsutil lifecycle set lifecycle.json gs://${BUCKET_NAME}/
Signed-URLs

First, create a Service Account, with just enough privileges to modify Cloud Storage, and add and download the assigned key.

$ gsutil cp test.txt gs://xtof-sandbox/
$ gsutil signurl -d 3m key.json gs://xtof-sandbox/test.txt

The above will return a signed-URL (it will look something like https://storage.googleapis.com/xtof-sandbox/test.txt?x-goog-signature=23asd...), which you can send to users and will only be valid for 3 minutes. After 3 minutes, they will get an "ExpiredToken" error.

Access control
$ gsutil cp foobar.txt gs://$BUCKET_NAME_1/
  • Get the default access list that has been assigned to foobar.txt:
$ gsutil acl get gs://$BUCKET_NAME_1/foobar.txt  > acl.txt
$ cat acl.txt
[
  {
    "entity": "project-owners-1045887948991",
    "projectTeam": {
      "projectNumber": "1045887948991",
      "team": "owners"
    },
    "role": "OWNER"
  },
  {
    "entity": "project-editors-1045887948991",
    "projectTeam": {
      "projectNumber": "1045887948991",
      "team": "editors"
    },
    "role": "OWNER"
  },
  {
    "entity": "project-viewers-1045887948991",
    "projectTeam": {
      "projectNumber": "1045887948991",
      "team": "viewers"
    },
    "role": "READER"
  },
  {
    "email": "storecore@sandbox-gcp-e2518d857e18d36a.iam.gserviceaccount.com",
    "entity": "user-storecore@sandbox-gcp-e2518d857e18d36a.iam.gserviceaccount.com",
    "role": "OWNER"
  }
]
  • Set the access list to private and verify the results:
$ gsutil acl set private gs://$BUCKET_NAME_1/foobar.txt
$ gsutil acl get gs://$BUCKET_NAME_1/foobar.txt > acl2.txt
$ cat acl2.txt
[
  {
    "email": "storecore@sandbox-gcp-e2518d857e18d36a.iam.gserviceaccount.com",
    "entity": "user-storecore@sandbox-gcp-e2518d857e18d36a.iam.gserviceaccount.com",
    "role": "OWNER"
  }
]
  • Update the access list to make the file publicly readable:
$ gsutil acl ch -u AllUsers:R gs://$BUCKET_NAME_1/foobar.txt
$ gsutil acl get gs://$BUCKET_NAME_1/foobar.txt > acl3.txt
$ cat acl3.txt
[
  {
    "email": "storecore@sandbox-gcp-e2518d857e18d36a.iam.gserviceaccount.com",
    "entity": "user-storecore@sandbox-gcp-e2518d857e18d36a.iam.gserviceaccount.com",
    "role": "OWNER"
  },
  {
    "entity": "allUsers",
    "role": "READER"
  }
]
Customer-supplied encryption keys (CSEK)
  • Generate a CSEK key (an AES-256 base-64 key):
$ python -c 'import base64; import os; print(base64.encodestring(os.urandom(32)))'
1G8l2isrv/QO5zJveoNCN5PeuGHgHBDUzHUBzgiOSUc=
  • Edit ~/.boto:
encryption_key=1G8l2isrv/QO5zJveoNCN5PeuGHgHBDUzHUBzgiOSUc=

Note: If the ~/.boto file is empty, generate it with:

$ gsutil config -n
$ gsutil rewrite -k gs://$BUCKET_NAME_1/foobar2.txt
Enable lifecycle management
$ gsutil lifecycle get gs://$BUCKET_NAME_1
gs://storecore-35635/ has no lifecycle configuration.
$ vi life.json
{
  "rule":
  [
    {
      "action": {"type": "Delete"},
      "condition": {"age": 31}
    }
  ]
}
$ gsutil lifecycle set life.json gs://$BUCKET_NAME_1
Setting lifecycle configuration on gs://storecore-35635/...
$ gsutil lifecycle get gs://$BUCKET_NAME_1
{"rule": [{"action": {"type": "Delete"}, "condition": {"age": 31}}]}
Enable versioning
$ gsutil versioning get gs://$BUCKET_NAME_1
gs://storecore-35635: Suspended

$ gsutil versioning set on gs://$BUCKET_NAME_1
Enabling versioning for gs://storecore-35635/...

$ gsutil versioning get gs://$BUCKET_NAME_1
gs://storecore-35635: Enabled

Upload a file, delete some lines from the original file, upload again, repeat.

  • List all versions of the file:
$ gsutil ls -a gs://$BUCKET_NAME_1/foobar.txt
gs://storecore-35635/foobar.txt#1552432005059570
gs://storecore-35635/foobar.txt#1552432853742567
gs://storecore-35635/foobar.txt#1552432873281759
Synchronize a directory to a bucket

Make a nested directory structure so that you can examine what happens when it is recursively copied to a bucket.

Run the following commands:

 $ mkdir firstlevel
 $ mkdir ./firstlevel/secondlevel
 $ cp foobar.txt firstlevel
 $ cp foobar.txt firstlevel/secondlevel
  • Sync the firstlevel directory on the VM with your bucket:
$ gsutil rsync -r ./firstlevel gs://$BUCKET_NAME_1/firstlevel
  • Verify that versioning was enabled:
$ gsutil versioning get gs://$BUCKET_NAME_1
$ gsutil cat gs://$BUCKET_NAME_2/$FILE_NAME

BigQuery

$ bq query "select string_field_10 as request, count(*) as requestcount from logdata.accesslog group by request order by requestcount desc"
+----------------------------------------+--------------+
|                request                 | requestcount |
+----------------------------------------+--------------+
| GET /store HTTP/1.0                    |       337293 |
| GET /index.html HTTP/1.0               |       336193 |
| GET /products HTTP/1.0                 |       280937 |
| GET /services HTTP/1.0                 |       169090 |
| GET /products/desserttoppings HTTP/1.0 |        56580 |
| GET /products/floorwaxes HTTP/1.0      |        56451 |
| GET /careers HTTP/1.0                  |        56412 |
| GET /services/turnipwinding HTTP/1.0   |        56401 |
| GET /services/spacetravel HTTP/1.0     |        56176 |
| GET /favicon.ico HTTP/1.0              |        55845 |
+----------------------------------------+--------------+

Cloud VPN

This section will show how to setup a VPN between two subnets in different regions.

Create VPC networks
  • VPC#1
    • Name: vpn-network-1
    • Subnet creation mode: Custom
    • Name: subnet-a
    • Region: us-east1
    • IP address range: 10.5.4.0/24
  • VPC#2
    • Name: vpn-network-2
    • Subnet creation mode: Custom
    • Name: subnet-b
    • Region: europe-west1
    • IP address range: 10.1.3.0/24
Create test VMs

One in both regions (VM#1: us-east1; VM#2: europe-west1)

Create firewall rules

VPC network -> Firewall rules

  • Allow traffic to vpn-network-1
    • Name: allow-icmp-ssh-network-1
    • Network: vpn-network-1
    • Targets: All instances in the network
    • Source filter: IP ranges
    • Source IP ranges: 0.0.0.0/0
    • Protocols and ports: Specified protocols and ports
      • tcp -> 22
      • other -> icmp
  • Allow traffic to vpn-network-2
    • Name: allow-icmp-ssh-network-2
    • Network: vpn-network-2
    • Targets: All instances in the network
    • Source filter: IP ranges
    • Source IP ranges: 0.0.0.0/0
    • Protocols and ports: Specified protocols and ports
      • tcp -> 22
      • other -> icmp
Create and prepare the VPN gateways

Create two VPN gateways, one in each region. Create forwarding rules for Encapsulating Security Payload (ESP), UDP:500, and UDP:4500 for each gateway.

  • Create vpn-1 gateway:
$ gcloud compute target-vpn-gateways \
    create vpn-1 \
    --network vpn-network-1  \
    --region us-east1
  • Create vpn-2 gateway:
$ gcloud compute target-vpn-gateways \
    create vpn-2 \
    --network vpn-network-2  \
    --region europe-west1
  • Reserve a static IP for each network:
$ gcloud compute addresses create --region us-east1 vpn-1-static-ip
$ gcloud compute addresses list
$ export STATIC_IP_VPN_1=<IP address for vpn-1>

$ gcloud compute addresses create --region europe-west1 vpn-2-static-ip
$ gcloud compute addresses list
$ export STATIC_IP_VPN_2=<IP address for vpn-2>
Create forwarding rules for both vpn gateways

The forwarding rules forward traffic arriving on the external IP to the VPN gateway. It connects them together. Create three forwarding rules for the protocols necessary for VPN.

  • Create ESP forwarding rule for vpn-1:
$ gcloud compute \
    forwarding-rules create vpn-1-esp \
    --region us-east1  \
    --ip-protocol ESP  \
    --address ${STATIC_IP_VPN_1} \
    --target-vpn-gateway vpn-1
  • Create ESP forwarding rule for vpn-2:
$ gcloud compute \
    forwarding-rules create vpn-2-esp \
    --region europe-west1  \
    --ip-protocol ESP  \
    --address ${STATIC_IP_VPN_2} \
    --target-vpn-gateway vpn-2
  • Create UDP500 forwarding for vpn-1:
$ gcloud compute \
    forwarding-rules create vpn-1-udp500  \
    --region us-east1 \
    --ip-protocol UDP \
    --ports 500 \
    --address ${STATIC_IP_VPN_1} \
    --target-vpn-gateway vpn-1
  • Create UDP500 forwarding for vpn-2:
$ gcloud compute \
    forwarding-rules create vpn-2-udp500  \
    --region europe-west1 \
    --ip-protocol UDP \
    --ports 500 \
    --address ${STATIC_IP_VPN_2} \
    --target-vpn-gateway vpn-2
  • Create UDP4500 forwarding for vpn-1:
$ gcloud compute \
    forwarding-rules create vpn-1-udp4500  \
    --region us-east1 \
    --ip-protocol UDP --ports 4500 \
    --address ${STATIC_IP_VPN_1} \
    --target-vpn-gateway vpn-1
  • Create UDP4500 forwarding for vpn-2:
$ gcloud compute \
    forwarding-rules create vpn-2-udp4500  \
    --region europe-west1 \
    --ip-protocol UDP --ports 4500 \
    --address ${STATIC_IP_VPN_2} \
    --target-vpn-gateway vpn-2
  • List forwarding rules (created above):
$ gcloud compute forwarding-rules list
NAME           REGION        IP_ADDRESS      IP_PROTOCOL  TARGET
vpn-2-esp      europe-west1  34.76.244.229   ESP          europe-west1/targetVpnGateways/vpn-2
vpn-2-udp4500  europe-west1  34.76.244.229   UDP          europe-west1/targetVpnGateways/vpn-2
vpn-2-udp500   europe-west1  34.76.244.229   UDP          europe-west1/targetVpnGateways/vpn-2
vpn-1-esp      us-east1      35.237.123.102  ESP          us-east1/targetVpnGateways/vpn-1
vpn-1-udp4500  us-east1      35.237.123.102  UDP          us-east1/targetVpnGateways/vpn-1
vpn-1-udp500   us-east1      35.237.123.102  UDP          us-east1/targetVpnGateways/vpn-1
  • List VPN gateways:
$ gcloud compute target-vpn-gateways list
NAME   NETWORK        REGION
vpn-2  vpn-network-2  europe-west1
vpn-1  vpn-network-1  us-east1
Create tunnels

Create the tunnels between the VPN gateways. After the tunnels exist, create a static route to enable traffic to be forwarded into the tunnel. If this is successful, you can ping a local VM in one location on its internal IP from a VM in a different location.

  • Create the tunnel for traffic from Network-1 to Network-2:
$ gcloud compute \
    vpn-tunnels create tunnel1to2  \
    --peer-address ${STATIC_IP_VPN_2} \
    --region us-east1 \
    --ike-version 2 \
    --shared-secret gcprocks \
    --target-vpn-gateway vpn-1 \
    --local-traffic-selector 0.0.0.0/0 \
    --remote-traffic-selector 0.0.0.0/0
  • Create the tunnel for traffic from Network-2 to Network-1:
$ gcloud compute \
    vpn-tunnels create tunnel2to1 \
    --peer-address ${STATIC_IP_VPN_1} \
    --region europe-west1 \
    --ike-version 2 \
    --shared-secret gcprocks \
    --target-vpn-gateway vpn-2 \
    --local-traffic-selector 0.0.0.0/0 \
    --remote-traffic-selector 0.0.0.0/0
  • List VPN tunnels (created above):
$ gcloud compute vpn-tunnels list
NAME        REGION        GATEWAY  PEER_ADDRESS
tunnel2to1  europe-west1  vpn-2    35.237.123.102
tunnel1to2  us-east1      vpn-1    34.76.244.229
Create static routes
  • Create a static route from Network-1 to Network-2:
$ gcloud compute  \
    routes create route1to2  \
    --network vpn-network-1 \
    --next-hop-vpn-tunnel tunnel1to2 \
    --next-hop-vpn-tunnel-region us-east1 \
    --destination-range 10.1.3.0/24
  • Create a static route from Network-2 to Network-1:
$ gcloud compute  \
    routes create route2to1  \
    --network vpn-network-2 \
    --next-hop-vpn-tunnel tunnel2to1 \
    --next-hop-vpn-tunnel-region europe-west1 \
    --destination-range 10.5.4.0/24
  • List static routes (created above):
$ gcloud compute routes list
NAME        NETWORK        DEST_RANGE     NEXT_HOP                            PRIORITY
...
route1to2   vpn-network-1  10.1.3.0/24    us-east1/vpnTunnels/tunnel1to2      1000
route2to1   vpn-network-2  10.5.4.0/24    europe-west1/vpnTunnels/tunnel2to1  1000
Verify VPN connectivity

SSH into VM1 and check that you can ping the internal IP of VM2, and vice versa.

SSH into instance/VM

SEE: gcloud compute ssh for details.

  • To SSH into "example-instance" in zone us-west2-a, run:
$ gcloud compute ssh example-instance --zone=us-west2-a
  • You can also run a command on the virtual machine. For example, to get a snapshot of the guest's process tree, run:
$ gcloud compute ssh example-instance --zone=us-west2-a --command="ps -ejH"
  • If you are using the Google Container-Optimized virtual machine image, you can SSH into one of your containers with:
$ gcloud compute ssh example-instance --zone=us-west2-a --container=CONTAINER
  • You can limit the allowed time to ssh. For example, to allow a key to be used through 2019:
$ gcloud compute ssh example-instance --zone=us-west2-a --ssh-key-expiration="2020-01-01T00:00:00:00Z"
  • Or alternatively, allow access for the next two minutes:
$ gcloud compute ssh example-instance --zone=us-west2-a --ssh-key-expire-after=2m

Get billing information

$ gcloud alpha billing accounts list
ACCOUNT_ID            NAME              OPEN   MASTER_ACCOUNT_ID
000000-000000-000000  Blue Env - Xtof   True   0A0A0A-0B0B0B-0C0C0C

$ gcloud alpha billing projects list --billing-account=000000-000000-000000
PROJECT_ID     BILLING_ACCOUNT_ID    BILLING_ENABLED
blue-123456    AAAAAA-BBBBBB-CCCCCC  True
green-654321   DDDDDD-EEEEEE-FFFFFF  False

GCP vs. AWS

Note: All of the following are as of February 2017.

GCP vs. AWS
GCP vs. Azure
  • Compute
    • Compute Engine vs. EC2
    • App Engine vs. Elastic Beanstalk
    • Container Engine vs. EC2
    • Container Registry vs. ECR
    • Cloud Functions vs. Lambda
  • Identity & Security
    • Cloud IAM vs. IAM
    • Cloud Resource Manager vs. n/a
    • Cloud Security Scanner vs. Inspector
    • Cloud Platform Security vs. n/a
  • Networking
    • Cloud Virtual Network vs. VPC
    • Cloud Load Balancing vs. ELB
    • Cloud CDN vs. CloudFront
    • Cloud Interconnect vs. Direct Connect
    • Cloud DNS vs. Route53
  • Storage and Databases
    • Cloud Storage vs. S3
    • Cloud Bigtable vs. DynamoDB
    • Cloud Datastore vs. SimpleDB
    • Cloud SQL vs. RDS
    • Persistent Disk vs. EBS
  • Big Data
    • BigQuery vs. Redshift
    • Cloud Dataflow vs. EMR
    • Cloud Dataproc vs. EMR
    • Cloud Datalab vs. n/a
    • Cloud Pub/Sub vs. Kinesis
    • Genomics vs. n/a
  • Machine Learning
    • Cloud Machine Learning vs. Machine Learning
    • Vision API vs. Rekognition
    • Speech API vs. Polly
    • Natural Language API vs. Lex
    • Translation API vs. n/a
    • Jobs API vs. n/a
  • Compute Services (GCP vs. AWS):
    • Infrastructure as a Service (IaaS): Compute Engine vs. EC2
    • Platform as a Service (PaaS): App Engine vs. Elastic Beanstalk
    • Containers as a Service: Container Engine vs. EC2
Compute IaaS comparison
Feature Amazon EC2 Compute Engine
Virtual machines Instances Instances
Machine images Amazon Machine Image (AMI) Image
Temporary virtual machines Spot instances Preemptible VMs
Firewall Security groups Compute Engine firewall rules
Automatic instance scaling Auto Scaling Compute Engine autoscaler
Local attached disk Ephemeral disk Local SSD
VM import Supported formats: RAW, OVA, VMDK, VHD Supported formats: AMI, RAW, VirtualBox
Deployment locality Zonal Zonal


Networking services comparison
Networking Load Balancing CDN On-premises connection DNS
AWS VPC ELB CloudFront Direct Connect Route53
GCP Cloud VirtualNetwork1 Cloud LoadBalancing2 Cloud CDN Cloud InterConnect Cloud DNS

1GCP allows for 802.1q tagging (aka VLAN taggin). AWS does not.
2GCP allows for cross-region load balancing. AWS does not.


Storage services comparison
Object Block Cold File
AWS S3 EBS1 Glacier EFS
GCP Cloud Storage Compute Engine Persistent Disks2 Cloud Storage Nearline ZFS/Avere


1An EBS volume can be attached to only one EC2 instance at a time. Can attach up to 40 disk volumes to a Linux instance. Available in only one region by default.
2GCP Persistent Disks in read-only mode can be attached to multiple instances simultaneously. Can attach up to 128 disk volumes. Snapshots are global and can be used in any region without additional operations or charges.


Database services comparison
RDMS NoSQL (key-value) NoSQL (indexed)
AWS RDS DynamoDB DynamoDB
GCP Cloud SQL1 Cloud Bigtable2 Cloud Datastore


1MySQL only.
2100 MB maximum item size. Does not support secondary indexes.


Big Data services comparison
Streaming data ingestion Streaming data processing Batch data processing Analytics
AWS Kinesis Kinesis EMR Redshift
GCP Cloud Pub/Sub Cloud Dataflow Cloud Dataflow / Cloud Dataproc BigQuery


Cloud Pub/Sub 
GCPs offering for data streaming and message queue. It allows for secure communication between applications and can also serve as a de-coupling method (a good way to scale).
Dataflow 
GCPs managed service offering for batch and streaming data processing. Apache Beam under-the-hood.
Dataproc 
GCPs offering for data processing using Apache Hadoop and Apache Spark. It is a massively parallel data processing and transformation engine.
Supported services: MapReduce, Apache Hive, Apache Pig, Apache Spark, Spark SQL, PySpark, and support for parallel jobs with YARN.
BigQuery 
GCPs offering for a fully managed, massive data warehousing and analytics engine, allowing for data analytics using SQL.
Application services comparison
Messaging
AWS SNS
GCP Cloud Pub/Sub


Cloud Pub/Sub (publisher/subscriber)
Management services comparison
Monitoring Deployment (IaC)
AWS CloudWatch CloudFormation
GCP Stackdriver Deployment Manager


See also

External links