OpenStack Enhances AI Workload Support for Improved Performance

OpenStack empowers enterprises to manage their own private clouds on-premises, resembling AWS infrastructure. After 29 releases, it remains one of the most active open-source projects globally. This week, the OpenInfra Foundation announced the launch of OpenStack version 29, known as “Caracal.” This latest release focuses on enhanced features specifically tailored for artificial intelligence (AI) and high-performance computing (HPC) workloads.

Large enterprises like Walmart and telecom giants like NTT are typical OpenStack users. Presently, nearly all enterprises are exploring how to operationalize their AI models while prioritizing data security. For many, this involves maintaining complete control over their entire technology stack.

As Nvidia CEO Jensen Huang recently pointed out, we are on the brink of a multi-trillion-dollar investment surge in data center infrastructure. While a significant portion is directed towards large hyperscalers, a considerable share will also support private deployments that require robust software management.

This trend positions OpenStack as a key player and one of the few comprehensive alternatives to VMware's offerings. VMware is currently facing challenges, with many users dissatisfied following its acquisition by Broadcom. OpenInfra Foundation executive director Jonathan Bryce noted that the situation has prompted numerous companies to reconsider OpenStack as an alternative since the Broadcom deal and recent licensing modifications.

Much of OpenStack's growth in recent years has stemmed from its adoption in the Asia-Pacific region. The OpenInfra Foundation recently welcomed Okestro, a South Korean cloud provider specializing in AI, as its newest Platinum Member. Europe, particularly the U.K. with its stringent data sovereignty regulations, has also become a significant growth market, as evidenced by the UK's Dawn AI supercomputer utilizing OpenStack.

“All indicators suggest a substantial rise in open-source infrastructure adoption,” remarked OpenInfra Foundation COO Mark Collier. “This trend encompasses OpenStack primarily, along with Kata Containers and other initiatives. It’s thrilling to witness another wave of infrastructure upgrades presenting our community with exciting challenges ahead for years to come.”

On a practical level, the new release features the capability for vGPU live migrations within Nova, OpenStack’s core compute service. This enhancement allows users to transfer GPU workloads between physical servers with minimal disruption, fulfilling a crucial request from enterprises eager to optimize the management of their expensive GPU resources. While live migration for CPUs has been a standard feature in Nova, this marks the first availability for GPUs.

Furthermore, this release brings significant security improvements, including rule-based access control for essential OpenStack services such as the Ironic bare-metal-as-a-service project. In addition to networking upgrades to optimize support for HPC workloads, the release includes a host of other enhancements. Comprehensive release notes can be found here.

This update marks the first since OpenStack adopted its “Skip Level Upgrade Release Process” (SLURP) last year. OpenStack typically rolls out a new release every six months, but this rapid pace has been challenging for most enterprises. In earlier phases, users often described the upgrade experience as “painful.”

Today, the upgrade process has been streamlined, resulting in a more stable project. The SLURP cadence introduces a long-term release format, where every alternate release each year is a SLURP version, simplifying upgrades while still providing major updates within the original six-month cycle for those seeking a quicker approach.

Over the years, OpenStack has experienced fluctuating perceptions but has evolved into a robust system supported by a sustainable ecosystem—a stark contrast to the tumultuous hype phase a decade ago. Recently, it has achieved notable success in the telecommunications sector, positioning itself well to harness the emerging opportunities from the AI revolution.