Photo by Niek Doup on Unsplash

Since open-sourcing Vespa.ai in 2017, it has been available as a container image for the x86_64 architecture. Vespa.ai is now released as a multiplatform image, supporting both x86_64 and ARM64. Read more at Vespa container images.

The Vespa Team has performance-tested ARM64 extensively to optimize code and validate performance. The CPU architectures have different performance and cost characteristics - application owners can now easily benchmark their applications on more alternatives to find the sweet spot for their workloads.

Supporting ARM64 gives application owners more options when selecting resources for deployment, and therefore more alternatives for vendor and cost optimization.

A common workflow for development using Vespa is developing code and configuration on a laptop. This is enabled by Vespa scaling from one to hundreds of nodes with no other configuration changes. The updated application is manually deployed to development/test systems for fast iteration, then checked in to a version control system for automatic production deployment. The multi-architecture image simplifies this flow - developers using e.g. Apple MacBooks with the M1 chipset can use the vespaengine/vespa image out of the box, regardless of deployment to other clusters.

ARM64 is also available on Vespa Cloud.