AWS Vastly Diversifies Its Portfolio Of VMs And Aims To Attract More Industries to Cloud Computing

Amazon Web Services (AWS) beefed up its cloud infrastructure portfolio Wednesday by introducing several virtual machine types intended to increase resource flexibility, delivering to customers a broader choice of compute environments for running their workloads.

The new instance families optimize for memory, compute or disk input/output speeds. AWS also introduced instances that enable hhardware acceleration through user-programmable processors, and advanced graphics processing capabilities that can be attached to any of its virtual machines.

AWS CEO Andy Jassy unveiled the new VMs in his keynote at the AWS re:Invent conference in Las Vegas, telling more than 30,000 attendees that the expanded selection will make more cloud projects practical and cost-effective.

[Related: AWS Channel Chief: New AWS Competencies Will Help Channel Partners Stand Out, Meet Customer Demand]

"It is so much easier to build whatever you want to build when you have the right tools at your disposal," Jassy said.

Complimenting those new configurations of CPU and other resources, Jassy introduced a service called Amazon Lightsail that simplifies launching virtual private clouds on Amazon's infrastructure. Lightsail eases AWS adoption for basic projects, he said, by allowing customers to choose an image and spin up virtual private servers, with storage and networking bundled, in a few easy clicks.

The new Elastic GPU category makes it possible to enhance all AWS instance types with access to graphics processors. Elastic GPU comes in two flavors: G2 instances that excel at graphics-intensive applications like video encoding, and the P2s for compute-intensive jobs like machine learning and high-performance databases that increasingly rely on the data-crunching capabilities of GPUs.

New memory optimized instances, R4s, are intended for hosting high-performance databases, distributed memory caches, and in-memory analytics for enterprise applications. They're 20 percent cheaper than the R3 instances they replace and immediately available, Jassy said.

At the start of next year, C5 instances will hit the market, running next-generation Intel Xeon processors that deliver high-compute performance for batch processing and distributed analytics functions often used in science, engineering and gaming applications.

Also available early next year will be a family of I3 instances best-suited for I/O-intensive workloads such as databases and transactional systems that frequently write to disk. They deliver up to a nine-fold improvement in IOPS over the previous generation, Jassy said.

The F1 family includes two new instances that are the first in the industry to offer hardware-acceleration capabilities by making it possible for customers to program Field Programmable Gate Arrays that can dramatically increase performance over general-purpose CPUs.

Jassy also told attendees that AWS has released two larger sizes of the popular T2 burstable instance types often used when hosting websites.

"I think it's fair to say we love ourselves some compute," Jassy said.

AWS now offers 10 compute instance types, in addition to serverless compute options, that are the backbone of a $13 billion business.

This kind of breadth and capability "is incredibly important as a company when you try to move all your workloads to the cloud," Jassy said.

Peter Castello, executive vice president of Boston-based TriCore Solutions, an AWS partner that focuses on enterprise application management, told CRN that Jassy's announcements essentially eliminated even more excuses for not taking advantage of the cloud.

"The reason people aren't moving is mostly fear, and that's where providers, companies like us, can help," Castello said.

Kevin Carroll, TriCore's director of solutions architecture, added, "it progressively seems like every hurdle that keeps you from moving to the cloud, they have a team for knocking down those hurdles."

Ben Butler, vice president of marketing at REAN Cloud, headquartered in Herndon, Va., said the hardware-acceleration option—the F1 instances that feature FPGA chips—might encourage some business sectors that have remained hesitant to adopt cloud computing to make the leap.

"I think it's going to unleash traditional industries that might not have chosen to go to the cloud like construction, manufacturing, electronics," he said.

Those industries often favor a different computing paradigm for their workloads perceived to be a better fit for mainframes, instead of virtual machines, Butler told CRN.

"It was almost a step function in the jump of features," Butler said of Amazon's VM portfolio upgrade, citing the nine-fold improvement in IOPS of the I3 family. "it wasn't incremental, they really boosted these instance types."

Even as AWS introduces more serverless compute capabilities like its Lambda service, they're still intensely focusing on upgrading traditional compute resources, and at a rapid clip, Butler said.

"Obviously compute still drives a lot of AWS usage and they keep innovating on that side," he told CRN.

REAN Cloud has seen customers that like the idea of adopting AWS, Butler said, but thought the instances available weren't well-suited for their specific workloads. But with Wednesday's product launches, that factor will dissipate.

That means partners like REAN have an even larger role to play in recommending appropriate instance types to their customers, and refactoring workloads for those new services, he said.