Next Nutanix is contemplating the day when it ports some of its wares to the Arm CPU family, but hasn't yet put the job on its to-do list and doesn't think its hypervisor needs to make the jump.

The Register put the possibility of an Arm port to CTO Manosiz Bhattacharyya since a primary theme of Nutanix's Next conference this week was its plan to spin out some parts of its stack as containerized services to run in cloud-native environments.

Hyperscalers like AWS and Oracle aren't shy about the fact their Arm-powered servers offer superior price/performance for some workloads when compared to their x86 fleets, the same economics may one day apply on-prem … so surely Nutanix needs to think about it?

Bhattacharyya has already done so.

He told The Register that the Nutanix Kubernetes Platform, announced this week, can run on bare metal and the company intends to make that possible on the x86 architecture and will consider Arm too.

Bhattacharyya said an Arm version will be contingent on whether major hardware vendors adopt the architecture "as a first-class platform."

When and if that happens, the CTO believes "there will be a lot of interest in the enterprise. But for now, x86 is all we see."

He's more certain that Nutanix's containerized services will one day be ported to Arm because the architecture is most often used to power the cloud-native applications on which the company is making huge bets.

"Bringing the container substrate to Arm makes sense and we will do it in the long term," Bhattacharyya told The Register. An Arm port of the AOS storage service that Nutanix has made its first hypervisor-free product is "absolutely doable in the long term," he added.

But a port of Nutanix's hypervisor is not on the agenda - Nutanix just doesn't see demand for VMs on Arm.

Bhattacharyya also noted that the Arm architecture needs attention due to Nvidia's use of it in the compute cores present in its GPUs. The presence of Arm cores in system-on-chips like Apple's homebrew silicon that put CPU, GPU, NPU, and memory on the same die, and enjoy performance improvements as bandwidth constraints that come with PCI aren't a factor, also have his attention.

The CTO expects to see more SoCs in more roles, and especially in AI. That's another field that Nutanix focused on at its Next conference this week, with the debut of a second GPT-in-a-box offering.

Bhattacharyya argued that the product is more than a bandwagon-jumping exercise because the stack of tools needed to run inferencing is diverse and complex, making the act of bundling it ready for swift deployment valuable - and deserving of a CTO's attention. ®

https://www.theregister.com//2024/05/23/nutanix_arm_server_plans/