Lots of attention is focused on interconnecting GPUs for AI clusters. But what effect does AI have on other segments of the network, such as storage clusters, East-West application flows or core-to-edge? Mansour Karam, VP of Data Center Products from Juniper, explains:
- The data center market is growing at a rapid pace, with AI/ML workloads contributing significantly to this growth. In four years, the market is projected to reach $32 billion, with AI/ML workloads growing at a rate of 46% year on year.
- The key to optimizing AI/ML workloads lies in the technology and infrastructure. Speeds and feeds are crucial, with 800 gig and 1.6 Tbps interfaces on the horizon. Innovations in optics and topologies are needed to minimize the cost and latency of these clusters.
- Beyond hardware, software plays a critical role in maximizing the efficiency of AI/ML workloads. Juniper's expertise in data feeds and speeds, combined with software from Apstra, can provide visibility into latency and performance, enabling automatic tuning of parameters for optimal job completion times.
