AWS launched the M6g Compute family in May which are based on the ARM Graviton2 processors and the R6g and C6g families in June. Since then the Graviton2 processors have been included in the RDS, EMR and ElastiCache services.
The main advantage of the 6g series is that it delivers better performance and is cheaper than the 5 and 5a series (x-86 Intel and AMD processor based respectively). For example, an R5a is about 10% cheaper than the R5 and the equivalent R6g is about 10% cheaper than a R5a instance. The performance is also up to 40% better than on an equivalent 5 series according to AWS and up to 65% from keydb.
For most businesses the main blocker to ARM based servers adoption is the cost of change of OS and app configuration to run on ARM rather than x86 processors, particularly for any commercial software solution. The savings will make it worthwhile for some people, but many customers will find the transition costs outweigh the benefits. If it was simple to migrate, customers could use Lambda serverless services rather than EC2 or container compute. However, for the open source RDS services (MySQL, PostGreSQL and MariaDB) and ElastiCache services, the underlying hardware is completely unimportant if the Platform service is exactly the same. Strategic Blue sees this as a significant change for cost and performance in Cloud services.
Intel’s delivery of the 10nm Ice Lake Server processors has reportedly been delayed for the third time until Q1 2021. This means that the launch of the Intel based M6,C6,R6 will also be delayed until some time in Q2 2021. AWS will take advantage of this delay by embedding their ARM based chips into even more services, which will make more efficient and cheaper services available.
Strategic Blue recommends using the 6g series in SaaS & PaaS as much as possible, with adoption of the EC2 instances where it makes sense for the business to do so.