The Solution
Most datacenter-hosted applications were not able to run active-active across two or more Availability Zones. To enable this refactoring was not an option due to costs, the limited timeframe available and in some cases, MN did not possess the source code.
Design
Xebia optimized the architecture by implementing resilience capabilities while working with the given limitations. That resulted in a solution with in-depth mitigations. For example: to recover from a single EC2 host failure it is not necessary to do a full failover to another datacenter. It has less impact (time to recover, changes to moving parts, etc) to relaunch the failed EC2 instance on a different EC2 host in the same region. The restart in the same AZ (Availability Zone) is often within minutes and requires no further changes or manual intervention.
Operations
This new solution design also requires new capabilities, processes and skills to operate the environment. Xebia worked together with the customer to adjust and improve their way of working with AWS. Resulting updated processes and procedures, resulting in a higher degree of automation and less manual intervention. MN staff is also trained by Xebia to run the daily operations and how to operate during a failover.
Recovery
To recover from a (rare) failure of an entire AZ, Xebia implemented AWS Backup for non-critical applications, where there is extensive time to recover in a second AZ. For critical applications, Xebia implemented a combination of making daily backups and continuous replication, using AWS Elastic Disaster Recovery Service (AWS DRS). The implemented technical controls are aligned with business needs to align costs versus requirements.
.png)
