Broadcom and Cisco tie AI factory deployment to workload needs

Broadcom and Cisco Systems are aligning AI factory deployment with specific workload demands instead of creating custom infrastructure for each case. Both companies have developed a validated design using VMware Cloud Foundation and Cisco Unified Computing System. This setup matches workloads with hardware, ranging from processors for smaller models to GPU-rich systems, as explained by Sabina Anja, Broadcom's chief technologist for the VMware Cloud Foundation Division.

Anja elaborates, "The whole idea behind the VMware Private AI Cloud was to provide customers with a solution to this complex systems challenge, allowing them to obtain a tailor-made solution that integrates top-of-the-line technology for their specific needs."

Anja and Jeff Nichols, Cisco's technical leader, discussed this approach during an exclusive broadcast at VMware Explore. They emphasized the importance of considering factors like latency, response time, and concurrent users when sizing AI deployments. Cisco's portfolio covers solutions like Unified Edge for local inference, AI POD configurations, and the powerful eight-GPU UCS C885A M8 for heavy-duty data-center workloads.

Nichols stated, "We start with the workload itself, not the system. So, what are the requirements? Is the end user going to be doing inferencing? Are they going to be doing retrieval-augmented generation, fine-tuning or training?"

The VMware Cloud Foundation virtualizes hardware, enabling customers to allocate processors or accelerators based on demand. With Cisco's validated systems, AI factory deployments can be fully configured out of the box, eliminating the need for continuous integration efforts. Nichols noted, "The philosophy behind the AI factory is that the end user doesn't have to be involved in the process, nor does the IT department need to act as an ongoing AI infrastructure architect or builder.

They become AI consumers. The AI factories arrive fully built and ready for deployment on day one and two." This setup ensures a smooth and efficient AI factory deployment experience.