Google Cloud targets high-performance workloads

Google Cloud has announced the preview availability of new Arm-based virtual machines, as well as a fully-managed job scheduling system called Batch.

The Arm VMs will form a new part of Google’s top-end Tau VM lineup, which are designed for scale-out workloads like large Java applications, web servers and media transcoding.

“Powered by Ampere Altra Arm-based processors, T2A VMs deliver exceptional single-threaded performance at a compelling price, making them ideal for scale-out, cloud-native workloads,” Google Cloud vice president and general manager of infrastructure Sachin Gupta said in a blog post.

The other announcement, Batch, is a fully managed job scheduling system designed for compute heavy applications. Batch jobs automatically manage their own compute resources, easing the burden on customers, and allowing them to run thousands of jobs with a single command.

“This improves the developer experience for executing HPC, AI/ML, and data processing workloads such as genomics sequencing, media rendering, financial risk modelling, and electronic design automation,” said Gupta, in the blog post.

Today's Infrastructure Spotlight event also highlighted several other recent moves, including price stabilisation for Google Cloud's Spot VM feature (which allows delay-tolerant workloads to be performed using Google’s idle machine cycles at a hefty discount from on-demand pricing) and the general availability of the company’s Cloud TPU v4 machine learning pods via a new ML hub.

The company also talked up new security features like Confidential VMs, where data is impossible to view, even with a debugger, from outside the trusted execution environment of the VM itself. 

There is also a new Cloud Intrusion Detection System, which uses machine learning to analyse threats and update itself on an automatic basis.