Graviton5-based EC2 R9g and R9gd instances reach general availability

AWS made EC2 R9g and R9gd instances generally available on August 31. The memory-optimized families use Graviton5 processors, with R9gd adding local NVMe-based SSD storage.

New capacity for memory-heavy services

The AWS launch post targets databases, caches, and analytics workloads and lists initial availability in Northern Virginia, Ohio, Oregon, and Frankfurt. It describes higher compute, memory, and networking capabilities than the preceding generation.

Those vendor comparisons give teams a reason to investigate, not a result for their own service. A cache with frequent network waits and a database constrained by storage may respond differently to the same processor change.

Our analysis: distinguish an Arm-to-Arm instance refresh from a move out of an x86 deployment. The latter should include a compatibility inventory for application binaries, libraries, operational agents, and container images before performance testing begins.

Keep the bottleneck visible

Select a representative workload and record the resource that currently limits useful work. Compare latency, throughput, and failure recovery under equivalent application settings, then inspect whether the limiting resource has moved.

If considering R9gd, decide explicitly which data can live on local storage and how the service reconstructs it after losing an instance. A faster cache is useful only when its rebuild procedure meets the application’s recovery expectations.

Measure cost per completed unit of work using the intended regional purchase model. Avoid turning an advertised percentage into a fleet-wide forecast without workload evidence. Start with a reversible canary for one service, preserve a compatible launch configuration for the previous family, and expand when the measured behavior supports the migration.