Google Cloud customers running high-performance applications can now launch compute instances with up to 96 virtual CPUs, thanks to an update released today. The new 96 vCPU instances can support up to 624GB of RAM, making them good candidates for applications that scale up to take advantage of additional compute capacity on a single node.
That’s important for customers running applications like in-memory databases and satellite image analysis. Descartes Labs, a startup that processes tons of imagery shot from above the earth, saw a 38 percent improvement in performance of their compression operations using the new machine types, according to Google’s blog post.
This move is part of a broader struggle among cloud providers to offer their customers high-performance computing environments. Tons of computing power comes with a matching price tag, which translates into major revenue for Google and its competitors, each of which is trying to grow its business as workloads shift to the cloud.
The new machine types are powered by Intel Xeon processors based on the company’s Skylake architecture. Google is providing customers with a set of Intel Performance libraries for building software that’s optimized to run on top of the more powerful chips, so they can get more out of the added compute capacity.
Right now, the new machine types are available through Google’s cloud regions in Oregon, Iowa, South Carolina, Montreal, Belgium, the Netherlands, Mumbai, Singapore, and Taiwan. The company plans to make them available more broadly across its global infrastructure in the future.