MMS • Renato Losio
Article originally posted on InfoQ. Visit InfoQ
AWS recently announced the general availability of the R6a instances, EC2 designed for memory-intensive workloads like SQL and NoSQL databases.. The new instances are built on the AWS Nitro System and are powered by AMD Milan processors.
With sizes from r6a.large to r6a.48xlarge, the new instances provide up to 192 vCPU and 1536 GiB of memory, twice the limit of the r5a, for vertical scaling of databases and in-memory workloads. Channy Yun, principal developer advocate at AWS, explains:
R6a instances, powered by 3rd Gen AMD EPYC processors are well suited for memory-intensive applications such as high-performance databases (relational databases, noSQL databases), distributed web scale in-memory caches (such as memcached, Redis), in-memory databases such as real-time big data analytics (such as Hadoop, Spark clusters), and other enterprise applications.
As covered recently on InfoQ, the Cockroach Labs’ 2022 cloud report reported that AMD-based instances running Milan processors are the price-for-performance leaders on the major cloud providers and a better choice than Intel ones. According to AWS, the new AMD memory-optimized instances provide up to 35% better compute price performance compared to the previous generation instances and cost 10% less than x86-based R6i instances.
Differently from the previous generation r5a, the r6a instances are SAP-certified for memory-intensive enterprise databases like SAP Business Suite. Mario de Felipe, global director at Syntax, notes:
The not great news is they are not HANA certified. Despite the R family at AWS being focused on DB workloads, and the R family is a great host of SAP databases (Oracle, SQLserver, DB2, or HANA), the HANA database runs on Intel (…) This makes the r6a family the best option available for SAP AnyDB option (excluding HANA).
R6a instances support 40 Gbps bandwidth to EBS in the largest size, more than doubling the R5a limit, and up to 50 Gbps networking. On the largest size, customers can also enable the Elastic Fabric Adapter, the network interface designed for running HPC and ML applications at scale relying on high levels of inter-node communications.
Supporting new AVX2 instructions for accelerating encryption and decryption algorithms, the AMD Milan processor provides Secure Encrypted Virtualization (SEV). Khawaja Shams, co-founder at momento, tweets:
So excited to see always-on memory encryption in the new R6a instances! Glad to see this support go beyond Graviton2/3 & the M6i instances.
The new instances are currently available in a subset of AWS regions: Ohio, Northern Virginia, Oregon, Mumbai, Frankfurt, and Ireland. The on-demand hourly rate goes from 0.1134 USD (r6a.large) to 10.8864 USD (r6a.48xlarge) in the US East regions.