12-09-2014 11:06 AM - edited 03-07-2019 09:49 PM
Folks - I am wondering just how much delay the encapsulation and de-encapsulation of a Fibre Channel (FC) frame into an Ethernet frame adds to the initiator-to-target session. In other words, is it reasonable to suggest that there are certain workloads that are so I/O intensive, that FCoE can act as a bottleneck or an inhibitor because of the added latency that it presents?
I certainly welcome comments from anyone who has thoughts around this, but I am ideally looking for someone who has specifically addressed this question and perhaps performed some benchmark testing.
Lastly, I understand that every workload has its own I/O profile and demands and that available BW (whether between the FCF and the storage target or on the converged links) plays a role. For now, let's just assume that BW is not a bottleneck and that, mathematically speaking, the available BW is comparable to what it would be in a native FC environment.
Thank you
12-09-2014 03:11 PM
Never done any testing with FC and FCoE, but here is a link to a test document with graph.
(page 14).
http://www.brocade.com/downloads/documents/research_and_test_reports/brocade-evaluator-group-fc-vs-fcoe-rt.pdf
HTH
12-10-2014 02:03 PM
Thanks, Reza. I actually know that document pretty well. its a bit of BS, as it stacks FC against UCS in a dishonest way, but there is some good info in there to extract.
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide