Intelligent digital transformation is driving change in both our personal and professional lives and the intelligence behind digital transformation is Artificial Intelligence (AI). Data centers running AI applications with massive amounts of data are recasting that data into pertinent timely information, automated human interactions, and refined decision making.
Intelligent lossless data center networks are the fabric that support AI applications using Remote Direct Memory Access (RDMA) for optimized communication. RDMA thrives on a lossless Ethernet fabric which in turn depends on Priority Based Flow Control (PFC) to ensure there is no packet loss. PFC is widely available but has known issues that prevent lossless data center networks from scaling and meeting performance objectives of today’s hyperscale cloud fabrics. One important issue with PFC is the possibility of creating deadlocks within the data center network.
In this session, learn about new industry standards that are being developed to automatically avoid PFC deadlocks and allow the Intelligent Data Center Network to scale beyond the bounds of current hand tuned high-performance computing (HPC) network fabrics.
The need to interact with the data center in real-time is more important than ever in today’s world where augmented reality, voice recognition, and contextual searching demand immediate results. Lossless data center networks must achieve cloud scale and deliver unprecedented levels of performance and reliability to meet these real-time demands. There is simply no occasion by which deadlocks can be allowed to occur in data center networks.