GCP - Compute Engine - part 1

 

Compute Engine:

Compute engine is a computing and hosting services that let you create and manage a infra structure. Google compute engine is the Infra structure as a Service [IaaS] components of Google cloud.

General purpose of Compute Engine:

E series is used for Dev and test environments. It is very efficient for lowest cost per core.

* Virtual Desktops

* Web and apps server with low latency

N series is a balanced and performance workloads. 

* CRM, BI or back office

* Data pipelines

* Databases

C series is used for high performance application.

* Game servers

* Ad servers

* Data analytics

* Media streaming and transcoding

* CPU based AI/ML

H series is a highest compute per cores.

* Game servers

* Media streaming and transcoding

* High performance computing (HPC)

* CPU based AI/ML

M series is highest memory per cores.

* Large database

* Modeling and simulation

Z series is a highest storage per cores.

* Data analytics

* Large horizontal database scale out

G series is an inference and visualization with GPUs

* Video transcoding

A series is a highest performance of GPUs

* Deep learning

* Recommendation Models

Preemptible Instances

  * Instances offered at a discount (60 to 91%) in periods of excess Compute Engine capacity

  * Compute Engine might stop these instance in case of need a more computes.

  * Run for a maximum of 24 hours

  * No SLA

Spot instances are similar like preemptible instances but it will extent after 24 hours.

Cloud Function [Platform as service]

  * Serverless light weight compute service

  * It will support the standalone function that respond to the events

  * It can be written using Java script, Python 3 or Java runtimes

Cloud Run: [Platform as Service]

  *  Container based  serverless platform

  * Request based auto scaling and scale to Zero

  * Built in traffic management

Cloud Storage:

Classified the cloud storage in GCP as follows:

Persistent Disks * Durable, high performance block storage for virtual machines * Performance scales with size of the disk and with number of vCPUs on the VM * Data stored redundantly Local SSD * High performance block storage for virtual machines. * Physically attached to the server * Higher throughput and low latency than persistent disks * Each local SSD is 375GB.

Choose the storage depends upon Database:

User case study of storage:

Cloud Storage - Un structure data [Videos, images, backups and achieve]

Persistent Disk - Disk for Virtual machines

Local SSD - Flash optimized databases, Hot caching layer for analytics

File store - Web content management, Rendering and media processing

Bigtable - High through put application such as Big data, IoT

BigQuery - Big data analytics, Business intelligent

Cloud Fire store - User profiles, cross device data synchronization