Make Sure the Network Isn’t the Problem! 400GE Considerations and Best Practices for Testing the Cluster Fabric
TimeTuesday, November 13th2:30pm - 3pm
Description400GE by its very nature, introduces new challenges—increased bit error rate, mandatory FEC, and the performance of the PCS in the presence of errors that 56Gb/s SERDES and PAM-4 encoding inherently provide. 400GE is destined to be adopted by the industry at least 2x faster than was 100GE, and is right around the corner for supercomputing clusters. With all new optics and cables, switches, and servers driven by 400/200/100/50/25GE speeds, the need for testing has never been greater. Testing not just by chip and equipment makers, but by independent benchmark and performance test organizations.
The end users—data centers, supercomputing environments, service providers, and network operators—need empirical data on which equipment will satisfy their particular network needs. This is evident in the use of traffic generation to verify SCinet circuits. It's a critical time to renew efforts to compare and benchmark the performance of these new network products and devices.