Fulcrum's FocalPoint FM4000: A Scalable, Low-Latency 10GigE Switch for High-Performance Data Centers

  • Authors:
  • Uri Cummings;Dan Daly;Rebecca Collins;Virat Agarwal;Fabrizio Petrini;Michael Perrone;Davide Pasetto

  • Affiliations:
  • -;-;-;-;-;-;-

  • Venue:
  • HOTI '09 Proceedings of the 2009 17th IEEE Symposium on High Performance Interconnects
  • Year:
  • 2009

Quantified Score

Hi-index 0.00

Visualization

Abstract

The convergence of different types of networks into a common datacenter infrastructure poses a superset challenge on the part of theunderlying component technology.IP networks are feature-rich,storage networks are lossless with controlled topologies, andtransaction networks are low-latency with low jitter, parallelmulticast.A successful Converged Enhanced Ethernet (CEE) switchshould pass the domain specific network tests, and demonstrate thesedisparate capabilities at the same time, while maintaining trafficseparation.The FocalPoint FM4000 Ethernet switch chip was designed andarchitected both to provide a rich Ethernet feature set and maintainthe highest performance around corner cases.It achieves thisthrough the use of a full-rate shared memory, parallel multicasting,switch architecture along with deeply pipelined frame processing.It implements traditional Ethernet, layer-3/4, and the new CEEfeatures.In this, paper we provide an extensive performanceevaluation of the FocalPoint FM4000 chip with a number of individualperformance tests including, port-to-port line rate and latency,fairness of flow control under N-to-1 hot-spot, and multicast linerate and latency tests.Finally, we explore the convergence bymeasuring the simultaneous performance of prioritized,flow-controlled unicast traffic and provisioned multicast trafficagainst the backdrop of full-rate best effort stressing traffic.The experimental results show that the FocalPoint FM4000 switchprovides an impressive flow-through latency of only 300 nanoseconds,which is insensitive to the packet size. The FM4000 delivers optimalperformance under hot-spot communication with a degree of fairnessabove 98\%, and provides an upper bound for latency in prioritizedmulticast, ranging from 1.2 to 4.3 microseconds, depending on theaverage size of the background best-effort traffic.A directcomparison with non-prioritized multicasts, shows a performancespeedup ranging from 29 to 38 times.