#High Availability

• High Availability means Designing For failure . We should design a system such that it is available in case of any failure .
•  Designing Multi Availability Zone (multi- AZ) :- It means deploying resources in more than one zone within same region . Each region has at least three AZ’s – geographically co-located & connected via low latency NW but isolated from each other .  
• Back-up and DR Strategy :- Based on RTO and RPO
• RTO :- Recovery Time Objective means within how much time since failure we want system back up and running as per accepted SLA .
• RPO :- Recovery Point Objective simply means time till where data can be recovered .
• DR strategy is about RTO (service up time) and RPO – point in time (data backup)
• RDS can be deployed as multi AZ – data replication is synchronous . If primary fails , with failover trigger , standby instance can become primary .
• If solution is deployed as multi Region , with high RPO and RTO snapshot of database can be taken and then copy to other region . Copy can take few hours . If shorter RPO , data migration service can be used to transfer data .
• Four main strategies - 1) Back-up & Restore 2) Pilot Lite 3) Warm standby & 4) Multi site Active - Active

1. #Back-up & Restore :- Cheapest , highest RPO & RTO , uses AWS Backup service , Can be used within Same Region Or Different Region

Below picture depicts backup across regions using #DLM (Data Lifecycle Manager Service)

· DLM is used to copy EBS vol snapshots stored in S3 to other region

· EFS can be restored to another region using AWS Backup service

· RDS can be restored to another region if we have implemented cross region automated backup which will be stored in S3 

2) #Pilot Light :-

• Data is restored using continuous Asynchronous Data Replication bet primary and DR region hence faster recovery than backup . Here all read replicas / clusters used from diff. region.
• Core infra already available (vpc , auto scaling group,(for load balancer) and aurora replica ) hence quick time to up services . These are always up.
•  Application Servers are switched off . Ready for DR.
•  Any changes made needs to be replicated in DR region in core infra. AWS CloudFormation service can help here.
•  Aurora and DynamoDb can have global database / table that can be available even from diff region without any data replication.

3) #Warm Stand-by :-

• It uses features of Back up / Restore as well as Pilot Light .
• The only diff. is it will have scale down resources of primary site in DR region (web and app servers also available in DR)
• Scale down means if primary has multi AZ then DR will have only single AZ with tier 1 and 2 (web and app server) 

4) #Multi Site Active / Active :-

• Lowest RPO & RTO
• High Cost
• Both sites are full scale and operational – multi region deployment
• No DR as such
• No failover triggered