Where can I find an explanation of Fibre Channel buffer credit, end-to-end flow control from the host HBA to the CU/DEV in a cascaded fabric switched topology? I am concerned about the buffer credit flow control issues in maintaining sustained throughput on DWDM 2 GB channels over a distance of 200KM.
Good question. I think Robert Kembel's book on switched fabrics has this covered, but I don't know for certain. Someone in your organization might have a copy.
Also, if you feel like reading dry stuff, the
FC-PH standards documents cover flow control in Section 26.
There are no end-to-end buffer credits in most
nets because most FC uses only Class 3. Class 2 is infrequently used, but it does have end- to-end buffer credits. I'm assuming you are using Class 2 based on your question.
According to the standard, both end-to-end and buffer credits are used with Class 2. When either is exhausted, the sending node stops.
B2B credits are exchanged over links -- E2E credits are exchanged between two nodes in a session.
The flow control scenario for FC is to blast away until you are out of credits. That's fine until you encounter a shared, constrained resource like your 2GB
There is nothing that manages flow control specifically for the DWDM link which is the thing you need to manage. However, I think the Fibre Channel Congestion Control (FCCC) feature in Cisco switches could conceivably do this. That assumes you have all Cisco switches – which you probably don't. FCCC works like FCN/BCN congestion control, if you are familiar with them.
Good luck tracking this down.
Editor's note: Do you agree with this expert's response? If you have more to share, post it in one of our