AWS Adds More AMD-Based Instances to Its Cloud
Amazon Web Services is widening its embrace of AMD’s Epyc server chips with a new batch of T3a instances powered by the processors and designed to handle workloads where the demand is more around bursting than compute capabilities.
AWS announced plans to roll out the instances late last year at its re:Invent user conference and this week is making the instances available. The company offered the first T1 instance in 2010 and eight years later rolled out the first of the T3 generation instances for workloads that at times need to burst due to rapid, temporary increases in demand. The new T3a instances are the latest offerings in this family of instances.
“These instances deliver burstable, cost-effective performance and are a great fit for workloads that do not need high sustained compute power but experience temporary spikes in usage,” Jeff Barr, chief evangelist for AWS, wrote in a blog. “You get a generous and assured baseline amount of processing power and the ability to transparently scale up to full core performance when you need more processing power, for as long as necessary.”
The ability to cost-effectively handle such bursty workloads is a basic reason for enterprises migrating to the cloud, according to Zeus Kerravala, principal analyst with ZK Research. In traditional data centers, businesses have had to design their infrastructures for the peak points of highest demand, which may only happen once or twice a year, leading to low hardware utilization rates.
“This is specifically designed for these burstable workloads where you don’t always need sustained amounts of compute power, but if you want it, it’s there,” Kerravala told Channel Futures. “You’re seeing much better price performance than it would be with traditional infrastructures.”
The instances also are built on AWS’ Nitro System, a project that the cloud provider has been building out for several years after announcing in 2017. Nitro includes such components as the Nitro Hypervisor and Nitro Card I/O Acceleration. The goal of the initiative has been to improve latency and performance of compute instances by offloading some of the networking, storage and management tasks onto the Nitro architecture.
AWS in March last year rolled out similar EC2 instances, the M5a and R5a, that also were built on the Nitro System and cost 10 percent less than comparable instances. At the Summit event in March, they introduced the M5ad and R5ad, which added block storage to the instances and both also running on Epyc processors.
Like other cloud service providers, AWS is looking to expand options available to customers, including in the processors they want to access, from AMD’s Epyc chips to GPUs from Nvidia to Intel CPUs. The company also has deployed homegrown, low-power Arm-based Graviton chips in certain EC2 instances, the fruits of AWS’ purchase of chip maker Annapurna Labs in 2015.
“Those are all designed for specific purposes,” ZK’s Kerravala said. “What Amazon is doing is giving customers a choice of whatever processor type they need in the cloud. So whatever type of workload that you’re running, Amazon has it. … It’s another service they can offer for a very specific type of workload.”
The broad array of offerings is an indication that AWS is listening to customers and address needs, he said.
“It shows that the cloud is not just for your run-of-the-mill workload,” Kerravala said. “Any kind you want to run now you can run in the cloud.”
The new instances come the same week that Amazon announced its latest quarterly financial numbers, showing that AWS continues to be a key driver of Amazon’s larger business. AWS revenue grew 41% over the same period last year, with sales increasing to …