Breaking News
Design How-To

Achieve higher InfiniBand performance with on-load processing

9/7/2010 03:28 PM EDT
1 Comment
More Related Links
View Comments: Newest First | Oldest First | Threaded View
User Rank
re: Achieve higher InfiniBand performance with on-load processing
docdivakar   9/21/2010 10:59:15 PM
Price and performance are the bottom line criteria for on-load vs. off-load choices in InfiniBand (IB). For transaction processing where roundtrip latency is a primary driver, one needs to look at the total cost of deploying either on-loaded or off-loaded setup and compare that against much lower cost version of Cat6A 10Gig fabric with off-load engines (TOE/POE). A good majority of the clusters are still using Cat6/6A fabric, often with no off-loading. If they can show matching performances in latency-sensitive applications, then it doesn't matter which version of IB is better! For throughput applications (like Ansys/CFD cited), it hardly matters whether your latency is a few milliseconds higher, it is more a bandwidth issue! Dr.MP Divakar

August Cartoon Caption Winner!
August Cartoon Caption Winner!
"All the King's horses and all the KIng's men gave up on Humpty, so they handed the problem off to Engineering."
Top Comments of the Week
Like Us on Facebook Parts Search

185 million searchable parts
(please enter a part number or hit search to begin)
EE Times on Twitter
EE Times Twitter Feed
Flash Poll
David Patterson, known for his pioneering research that led to RAID, clusters and more, is part of a team at UC Berkeley that recently made its RISC-V processor architecture an open source hardware offering. We talk with Patterson and one of his colleagues behind the effort about the opportunities they see, what new kinds of designs they hope to enable and what it means for today’s commercial processor giants such as Intel, ARM and Imagination Technologies.