Google adds new family of VMs to its Compute Engine

Google adds new family of VMs to its Compute Engine

Google has added a new kind of virtual machine to its Compute Engine, promising reliable performance, flexible configurations, and “best total cost of ownership” for general-purpose workloads.

The new VM family runs under the E2 name and is now available as a beta offering in the company’s Iowa, South Carolina, Oregon, Northern Virginia, Belgium, Netherlands, Taiwan, and Singapore regions. Users can choose between custom and fifteen predefined configurations, which vary in terms of CPUs (16 max) and memory (up to 128GB). Pricing depends on the selected region, the machine type and the duration users want to commit to.

Google promises that its VMs can sustain a high CPU load without the throttling customers of other providers may encounter. Compared to other machine types available on GCP, E2 doesn’t support GPUs, local SSDs, sole tenant nodes, or nested virtualisation, which is something to keep in mind.

According to the official announcement, E2 is meant to save users 31 per cent of what they’d pay for a comparable N1 configuration while getting a similar performance. It is therefore promoted as an alternative for customers whose workloads do well on N1 but don’t depend on GPUs, local SSDs, or need large instances to work. Example use cases span scenarios like web or app serving, development environments, back office applications, or virtual desktops.

The lower costs seem to be mainly down to the straight forward architecture and resource balancing E2 uses, which is said to improve hardware usage in the company’s own latency-crtitical services as well. For smaller workloads like microservices special shared-core instances are available for additional savings.

Under the hood, the E2 VMs use Intel Xeon processors with AMD EPYC soon to come, which are automatically selected based on availability. Details can be found in the product’s documentation.