Google LLC’s cloud business today introduced a new series of instances based on Arm Ltd. silicon and a managed service that promises to simplify batch processing tasks. 

Both offerings made their debut at the company’s Infrastructure Spotlight event.

The latest addition to Google Cloud’s instance portfolio is the Tau T2A instance series. The series is the first from the company to be powered by Arm-based processors. In particular, Tau T2A instances are powered by processors from Ampere Computing LLC’s Arm-based Altra chip family. 

Ampere is a chip design startup backed by about $426 million in funding. Microsoft Corp., one of Google’s major competitors in the cloud market, also uses the startup’s Altra chips. 

Google said the Altra processors powering its new Tau T2A instances are based on Arm’s Neoverse N1 central processing unit design. Introduced in 2019, Neoverse N1 can run some server workloads up to 2.5 times faster than its predecessor. It provides at least 60% higher speeds than Arm’s previous-generation CPU design when performing processing tasks that involve integers, a basic unit of data.

The speed of Ampere’s Arm-powered Altra chips will translate into increased performance for Google Cloud customers. In a blog post, Ampere stated that the new Tau T2A instances can outperform virtual machines running on Intel Corp. silicon by up to 31%. Moreover, the startup is promising up to 65% better price-performance in certain situations.

Tau T2A instances can be configured with up to 48 virtual CPUs. An organization can provision up to 4 gigabytes of memory and 32 gigabits per second of network bandwidth for every instance that it uses. At the software level, meanwhile, the Tau T2A series supports several popular Linux distributions including Red Hat’s widely used RHEL platform.

“T2A VMs deliver exceptional single-threaded performance at a compelling price, making them ideal for scale-out, cloud-native workloads,” Google Cloud product managers Subra Chandramouli and Jamie Kinney detailed in a blog post. “Developers now have the option of choosing the optimal architecture to test, develop and run their workloads.”

Google Cloud has integrated the Tau T2A instance series with several of the other services in its platform. The Google Kubernetes Engine service, for example, now enables users to run their software container workloads on the new instances. The Tau T2A series also works with Google Cloud Batch, a new managed service that the search giant launched into preview this morning.

Some workloads, such as e-commerce applications, are expected to process a large number of user requests in parallel. Other workloads process far fewer user requests. A data migration tool, for example, can transfer information between two systems without requiring active input from an administrator while the information is being moved.

Workloads that run with little to no user input are commonly known as batch jobs. The new Google Cloud Batch service that launched into preview today is designed to ease the management of such workloads.

The service is delivered on a fully managed basis. As a result, administrators don’t have to maintain the underlying infrastructure or perform related maintenance tasks. Google Cloud Batch also includes relatively simple controls that make it possible to run thousands of batch jobs with a single command, according to the search giant. 

Google Cloud envisions customers using the service for a variety of tasks. The service is capable of running artificial intelligence models, chip design software and many other types of applications. Google Cloud Batch also supports MPI libraries, a set of specialized software tools used to run workloads that split computations across multiple chips to speed up processing. 

Image: Google