Starting this week customers of Google's Compute Engine infrastructure-as-a-service platform will be able to run their workloads on cloud servers based on Intel Corp's next generation Skylake processors.
Skylake-based virtual machine (VM) instances can support up to 64 virtual CPUs and 455GB of memory. They are designed to run a range of highly compute-intensive workloads such as genomic research, 3D rendering and data analytics.
Google announced beta availability of Skylake on Compute Engine in February. Since then organizations have run millions of hours worth of workloads on VMs based on the technology, Google product managers Hanan Youssef, Sami Iqram and Scott Van Woudenberg announced in a blog this week.
Google, like other cloud providers, allows customers to create a variety of virtual machine types to suit their requirements. Supported options include high memory configurations, high CPU types, custom machine types and pre-emptible VMs. With this week's announcement, enterprises will have the option of provisioning any of these VM instance types on Skylake-based hardware, the three product managers said.
For the next 60-days enterprises interested in the technology will be able to spin up Skylake VMs at no additional cost. After the end of the promotional period, enterprises that want Skylake VMs will pay a 6 percent to 10 percent premium compared to running VMs on previous generation Intel hardware.
Skylake based VMs are currently available to Google's cloud customers in the Western U.S., Western Europe and Eastern Asia Pacific cloud regions. Google will eventually make the technology available to customers in other regions as well.
"Customer demand for Skylake has been very strong," Youssef, Iqram and Van Woudenberg said. "We have more capacity arriving every day, and support for additional regions and zones coming in the near future."
In addition to the Skylake rollout, Google this week announced other Compute Engine updates that are designed to give organizations more choice and flexibility in running VMs on the company's cloud infrastructure.
Starting this week Google has removed memory caps for virtual machine instances. Instead of the previous 6.5 GB limit for a virtual CPU, organizations now have a maximum of 455GB of memory per VM. This will give organizations more flexibility for running applications like in-memory databases and high-performance relational databases on Google's cloud infrastructure.
Also available starting this week are more virtual machine type options, and an easier process for selecting a baseline processor type for a VM. Google's cloud platform systems are currently based on a variety of Intel processors including Sandy Bridge, Ivy Bridge and Broadwell. Each of the systems supports unique features and capabilities.
With this week's update, Google has made it easier for organizations to specify the minimum, or baseline, CPU platform for their virtual machines. The company's new Minimum CPU Platform feature lets organizations specify the CPU platform they want in a particular cloud zone and Google's Compute Engine will then always schedule that CPU family or above for running the customer's workload, the Google product managers said.