How Datadog migrated its Kubernetes fleet on AWS to Arm at scale

Matthieu Jaillais and Aaron Kaplan from Datadog share their experience of migrating nearly their entire Kubernetes fleet on AWS to Graviton-powered EC2 instances, a move that brought about significant cost savings and improved resilience. The migration process was complex, requiring careful planning, benchmarking performance, monitoring deployments in staging and production, and iterating as necessary to optimize. To track the migration, Datadog defined four key performance indicators (KPIs) - Arm adoption rate in production, baseline Arm-readiness, share of exceptions, and Jira tracking coverage. These KPIs were used to create two dashboards: one for engineers and another for executives, providing visibility into the migration's status and progress. The migration resulted in a 10% reduction in AWS bill and improved flexibility, durability, and failover options, paving the way for future multi-architecture opportunities with other cloud providers.