Resilient Architectures - Availability

Question 1

Redundancy, loose coupling and availability

Answer 1

• Redundancy enables loose coupling, and this improves availability and performance (by allowing to scale services independently)

Question 2

How to calculate availability?

Answer 2

It’s calculated separately by each service:

• Between different AWS services, or hard dependencies (if that service goes down the application becomes unavailable):
  1. Multiply the availability of components together
• Between the same services, or redundant components:
  1. Multiply the failure rate of every component
  2. Subtract that number from 100%

NOTE: Decide your availability requirements before beginning your design

Question 3

Loose coupling

Answer 3

• Means when one component doesn’t depend on a specific component
• Elastic services are always composed of redundant components, and always loosely coupled with other resources
• SQS is useful to connect loosely coupled services
• When using ELB there is a one-to-many relationship with the clients

Question 4

Cloud native applications

Answer 4

• Depend on a cloud service that can’t be deployed on on-premises
• Availability not necessarily higher than traditional applications, so it can be improved by using multiple regions
• Trusted Advisor can be used to find out your service limits