Compute in the cloud

1. ec2

2. • instance types

   • pricing

   • dedicated hardware

3. scaling

4. • auto-scaling

   • load balancing

5. messaging

6. • Simple Notification Service (SNS)

   • Simple Queue Service (SQS)

7. Serverless

8. • Lambda

9. container services

10. • Elastic Container Service (ECS)

    • Elastic Kubernetes Service (EKS)

    • Fargate "serverless containers"

(1) EC2

EC2 instance types

1. general purpose - balances compute, memory, network

2. compute optimized - high-performance processors

3. memory optimized - high-performance databases (db needs memory)

4. storage optimized - data warehousing

5. accelerated - (special use-cases)

R6a

R<version><cpu-arch>

6-> v6

a amd

i intel

EC2 Pricing

pay for what you use

EC2 instance pricing options

1. on-demand (pay for what you use)

2. spot (cheaper, might be interrupted)

3. reserved (cheaper, pay for instances even when not using it)

4. compute savings plan (cheaper, pay for consistent compute usage)

Dedicated Hardware (vs shared hardware with virtualization)

• dedicated instance (eg. legal compliance)

• dedicated host - physical box (eg. licensing requirements)

(2) scalability

auto-scaling

• get more instances when (eg. 3 minutes of 70% usage)

elastic load balancing

• get the traffic into the instance

side-effect: resilience when an instance crashes

(3) AWS messaging services

Simple Notification Service (SNS)

push / publish

topic

subscriber

Simple Queue Service (SQS)

queue

stored on queue

pull

(4) Serverless compute services

Lambda

(5) container services

Elastic Container Service (ECS)

Elastic Kubernetes Service (EKS)

Fargate - "serverless containers"

fargate pricing (per vCPU / GB per hour)

https://aws.amazon.com/ecs/pricing/