Introduction

Virutal machine in the cloud

Resizable compute capacity in cloud

Reduced time required to obtain and boot new server instances to minutes

Pricing

Pay as go, pay for what you use, pay less as you use more and pay even less when you reserve capacity.

On Demand

  • Pay fixed rate by the hour
  • Short term, spiky and unpredicatable workloads8

Reserved

  • Steady use and predictable usage
  • Provides capacity reservation, and significant discount on hourly charge. Contract terms are 1 year to 3 year terms
  • Standard Reserved instances
  • Convertible Reserved instances provides ability to convert between different instances
  • Scheduled Reserved instance provides reserving for certain times

Spot

  • Flexible stat and end times
  • Bid whatever price you want for instance capacity.
  • Fesible at very low compute instances

Dedicated Hosts

  • Used in Server-bound licenses
  • Regulatory requirements
  • Can be purchased On-demand
  • Can be purchased as reservation

EC2 Instance Types

The numbers are Generation of the family

Mnemonic → FIGHT-DR-MCPXZ-AU

Security Group

Inbound & outbound rules are stateful. Inbound rules get copied over to outbound rule

Security group takes effect immediately

Can’t block individual IP with security group

All ports are blocked by default

Network Access control list is stateless whereas security groups are stateful

Can’t block individual port with security group

Multiple security group can be attached to a EC2 instance

Multiple EC2 instances in a security group.

EBS Volumes

Elastic Block storage volumes replicated within availability zone to protect from component failure

5 different EBS Storage

General Purpose (SSD) → gp2

Provisioned IOPS SSD → io1

Throughput Optimized HDD → st1

Cold HDD → sc1

EBS Magnetic → Standard

EBS Volumes will be same Availability zone as EC2

EBS root Volume Killed when EC2 instance get killed because of delete on termination is enabled by default.

EBS Volumes can be modified on the Fly even when EC2 is running

EBS volumes can be moved to another availability zone by taking snapshot

that will create a image and it is used to create a volume in another availability zone

Para Virtual Virtualization vs Hardware virtualization.

Snapshot AMI can be copied to a different Region.

Addditonal Volumes that are not deleted when EC2 instance is killed.

Snapshot exist on S3

Snapshots are incremental

Its best practice to stop Ec2 instance before taking snapshot but its not required.

AMI Types

Instance Store (EPHEMERAL STORAGE)

→ cannot be stopped. Can only be terminated or rebooted

→ When stopped , the underlying host fails, you will lose data

→ No option to keep root device upon termination

EBS Backed

→ Can be stopped

→ Can tell EBS to keep the root device volume

ENI vs ENA vs EFA

ENI → Elastic Network Interface

Simple virtual network card

Basic networking

EN → Enhanced Networking

Single root I/O virtualization

Speed between 10 GBPs and 100 GBPS

EFA → Elastic Fabric Adapter

Accelrate High performance Computinng(HPC)

Can use OS-bypass , enable HPC

Encrypted Root Device Volumes

Create a AMI Create a copy of the snapshot

Create a image from the snapshot that is encrypted

Launch a new ec2 instance with the encrypted image.

Spot Instances &Spot Fleets

Take adavantage of unused EC2 capacity in the AWS cloud. Its good for stateless applications.

Hourly spot prices

Spot block can be setup form stop instance being terminated in 2minutes

Good for Big data and analytics, containerized workloads, CI/CD and testing,[

web services, Image and media. High-performance computing

Not good for Persistent workloads Critical jobs and databases

Spot Fleets are collection of Spot instances, optionally on-demand instances

EC2 Hibbernate

Hibernation save the contents from the instance memory to EBS volume

Stoppinng an EC2 instance ram is cleared.

Instance boots much faster,

Useful for long running process

Root device volume have to be encrypted to enable hibernation.

EC2 Placement group

Cluster Placement group

Low Network Latency/ High Network Throughput

Can’t span multiple zones

Spread Placement group

Individual Critical EC2 Instances

Can span multiple zones

Partition Placement group

Multuple EC2 instances HDFS

Homogenous instance with cluster placement group highly recommended