1. GCP Services

GCP includes IaaS and PaaS. SaaS, such as Gmail, fall outside of the scope of GCP.

IaaS

Takes care of the hardware level.
Requires more overhead as you're responsible for creating and managing instances etc.
Pay only for what you use.

Compute Engine
- raw computing
- granular control
Container Engine
- datacenter as computer
- declarative management

PaaS

Takes care of both the hardware and software, e.g. runtimes, dependencies, redundancy etc.
Write code and go - focus on app logic by making use of pre-set runtimes.
Pay only for what you use.

App Engine
- pre-set runtimes
- focus on app logic
Cloud Endpoints
Manage VMs
- bring your own runtime
- health-checked VMs

Compute Services

Compute Engine

Sort of like building a real pc, but through an API instead of physical components.

run large-scale workloads on Google's VMs
robust networking features
instance metadata & startup scripts
high CPU, high memory, standard, & shared-core machines
persistent disk snapshots
APIs for auto-scaling & group management

Container Engine

Container is Kubernetes and Docker

based on Kubernetes
orchestrate & schedule docker containers
consumes Compute Engine instances
uses declarative syntax to manage applications
manages and maintains:
- logging
- health management
- monitoring
- scaling

App Engine

Focus on features rather than infrastucture.

managed runtimes for specific versions of python, node, Go, etc.
autoscale web workloads to meet demand
free daily quotas and usage-based pricing
local SDK for dev and deployment
must conform to sandbox constraints:
- no writing to local fs
- request timeouts at 60s
- limit on 3rd-party software installations

Storage Services

Cloud Bigtable

NoSQL db service for web, mobile, and large-workload apps
- starts at terabyte, scales up to petabyte range
integrated
- accessed using Apache HBase API
- native comppatbility with big data ecosystem - let Google do the heavy lifting
protected
- replicated storage
- data encryption
proven
- drives many major Google apps

CLoud Storage

Useful for storing backups and archiving data - where most people get started.

high performance immutable blob storage
not a file system (Fuse can be used to treat it as such)
simple administration and does not require capacity management
all storage options accessed through the same APIs and include client libs:
- JSON API
- XML API

Cloud SQL

Fully managed SQL databases

Google-managed SQL
pay-per-use model
REST API for management
used for affordability and performance
Google security
Vertical scaling (reading and writing)
Horizontal scaling (reading)
seamless integration with GAE and GCE

Cloud Datastore

NoSQL store for billions of rows
Schemaless access - no need to think about underlying data structure
local dev tools
auto-scaling and fully managed
built-in redundancy
support for ACID transactions
includes free daily quota
access from anywhere via RESTful interface

Bigtable vs Datastore

Bigtable - what it's good for

Cloud Datastore - what it's good for

Big Data

BigQuery

near real-time interactive analysis of massive datasets (Tb's of data in seconds)
columnar storage for high performance
queries using SQL-like syntax
only pay for storage and processing
zero admin for perf and scale
supports open standards

Cloud Pub/Sub

scalable and reliable messaging
supports many-to-many async messaging
includes support for offline consumers
based on proven Google tech
integrates with Cloud Dataflow for data processing pipelines

Cloud Dataflow

construct scalable and reliable data pipelines
executes data processing on GCE instances
provices support for
- ETL
- analytics
- real-time computations
- process orchestration
integrates with GCP services
- Cloud Storage
- Cloud Pub/Sub
- BigQuery

Cloud Datalab

Get insights from your data

interactive tool for large-scale data exploration, transformation, analysis, and visualisation
- analyze data in GBQ, GCE, GCS
- easily deploy analyis models for GBQ
integrated and open source
- runs on App Engine
- built on Jupiter
- use Google Charts for easy visualisations
code, docs, results, and visualisations combine in an intuitive notebook format

Cloud Dataproc

fast, easy, managed way to run Hadoop and Spark on GCP
benefit from cloud integration
- Cloud Storage
- Cloud Monitoring
- Cloud Logging
Dataproc clusters are billed minute-by-minute
any-time scaling - manually scale clusters even while jobs are running
RESTful API and Cloud SDK integration

App Services

Cloud Endpoints

build your own API running on App Engine
expose your API using a RESTful interface
includes support for OAuth 2.0
Generate client libraries
includes some App Engine features:
- scaling
- denial of service protection
- high availability
supports iOS, Android, and JS clients

Translate API

translate arbitrary string between thousands of language pairs
programmatically detect a language
support for many languages
supports standard Google API client libs

Prediction API

pattern-matching and machine-learning
predict trends based on historical data
use cases include spam detection and product recommendations
data replicated using Cloud Storage
integrates with other GCP services
- App Engine
- BigQuery
- Cloud Storage

Networking

Cloud DNS

create and manage DNS programmatically

HTTP(S) Load Balancing

balance HTTP-based traffic across multiple GCE regions
simple DNS setup
scalable - requires no pre-warming

Network Load Balancing

spread TCP & UDP traffic over a pool of instances within a region
only healthy instances handle traffic
scalable, requires no pre-warming

Cloud Management Services

Cloud Monitoring

dashboards and alerts for cloud apps
review metrics fro cloud services, VMs, and commong open source apps

Cloud Launcher

launch software packages on GCE in seconds
WordPress, LAMP, Jenkins etc.

Cloud Deployment Manager

Declarative management

infrastructure management service
create a template describing your environment and use deployment manager to create resources

Cloud Logging

collect and store logs from GCE and GAE
view logs with log viewer
export logs to GCS and GBQ