Penguin Solutions revamps its software to automate mass-scale AI infrastructure deployments

Artificial intelligence infrastructure management company Penguin Solutions Inc. said today it’s expanding its flagship ICE ClusterWare platform so customers can use it as a foundation for building “fully optimized AI ecosystems” that can scale up as they grow.

New features include multi tenancy support and streamlined workflows, and a new ICE ClusterWare AIM service, which is said to help optimize the infrastructure using “predictive automation” to ensure it delivers maximum performance and availability

Penguin, which was previously known as Smart Global Holdings until it rebranded last year, boasts a long, multidecade history of working with high-performance computing systems, and has recently pivoted to focus exclusively on AI infrastructure. It helps customers design, build, deploy and manage the advanced computing and networking hardware needed to run AI workloads on-premises.

Penguin Vice President of Software and Product Management Trey Layton said that running AI infrastructure necessitates a more integrated and intelligent approach compared with traditional information technology systems. “We enable enterprises to build intelligent compute environments that optimize efficiency, scalability, and cost-effectiveness, ensuring AI workloads run at peak performance while minimizing operational complexity,” he said.

The ICE ClusterWare platform is the company’s main tool responsible for that. It’s a hardware-agnostic intelligent software platform for managing, scaling and optimizing AI infrastructures that combines various open-source and industry-standard tools with some of the company’s own proprietary offerings. Using the platform, companies can create “Intelligent Compute Environments” that can scale in tandem with the growth of their AI applications.

The latest updates throw in multitenancy, which enables secure and logical resource isolation, complete with dynamic workload partitioning to support multiple users and departments that require their own, separate AI environments.

Other updates include enhanced orchestration features, including a unified control plane for managing sub-clusters, and streamlined workflows, which are said to simplify AI infrastructure deployment and scaling with automated policy enforcement and job scheduling.

As for ICE ClusterWare AIM, this leverages AI algorithms that have been trained on more than 2 billion hours of graphics processing unit runtime. It’s meant to help simplify day-to-day operations in the data center, providing capabilities such as automated remediation to resolve system inefficiencies and ensure optimal performance, and prescriptive maintenance, which helps to mitigate system failures before they occur, reducing downtime. In addition, it helps to maximize GPU utilization by optimizing workload distribution, helping companies to improve their return on their AI infrastructure investments.

International Data Corp. analyst Ashish Nadkarni said that last feature should be especially useful, as traditional approaches often result in GPUs being massively underutilized, running at just 50% of their potential value on average.

“Without the right tools and expertise, AI infrastructure quickly becomes expensive and difficult to scale,” Nadkarni added. “Penguin Solutions’ ICE ClusterWare software and AIM services provide organizations with an integrated, automated approach to AI infrastructure, making it easier to deploy, manage, and optimize high-performance environments.”

Image: SiliconANGLE/Meta AI