Site icon AIT365

Mirantis OpenStack for Kubernetes Improves Support for AI and Windows Workloads

Mirantis

Mirantis announced the latest release of Mirantis OpenStack for Kubernetes (MOSK). The release incorporates hundreds of improvements from the open source upstream, plus dozens of innovations including graphics processing unit (GPU) virtualization – especially useful for cost-effectively running artificial intelligence (AI) workloads.

MOSK 24.1 introduces GPU virtualization as a technical preview, letting multiple cloud applications efficiently share resources of physical NVIDIA GPU boards. GPU virtualization helps optimize costs – maximizing utilization of expensive GPU hardware – while letting more teams and business units take advantage of processing power. Leveraging NVIDIA’s AI Enterprise software platform, MOSK enables physical processing resources to be portioned and presented to workloads as dedicated resources for accelerated performance and optimal resource allocation.

“In today’s rapidly evolving landscape, organizations are increasingly reliant on AI and highly-intensive compute tasks for competitive advantage,” said Roman Zhnichkov, director of engineering, Mirantis. “MOSK 24.1 empowers enterprises to harness the power of NVIDIA GPU virtualization technologies, enabling cost-effective AI hosting while enhancing operational agility.”

Also Read: Intel Announces New Edge Platform for Scaling AI Applications

MOSK 24.1 is built on OpenStack Antelope – the first OpenStack release designated as SLURP (Skip Level Upgrade Release Process). Users of MOSK 24.1 running OpenStack Yoga will be able to upgrade directly to Antelope, skipping the intermediate OpenStack Zed. This saves operational overhead for users to stay current. Other key improvements include:

MOSK enables on-premise cloud infrastructure for cloud-native and traditional applications, ensuring reliability and complete control over application data. MOSK provides automated management of the underlying infrastructure, from hardware provisioning to software configuration, plus centralized logging, monitoring, and alerting.

SOURCE: BusinessWire

Exit mobile version