FC Latency

Hi Guys,

I am wondering if someone can help me with a query - Consider a network has Fabric-A and B. Suppose a server is configured to send FC traffic over both HBAs in a round-robin fashion i.e. it sends a 1000 frames over Fabric-A and then another 1000 frames over Fabric-B. As I understand, Fabric-A will acknowledge to the server BEFORE server will start sending another 1000 frames over Fabric-B. 

Would this be the case, or am I missing smth? 

Also, how would the Storage controller react to frames arriving to it if, say, the path to Fabric-B is higher cost?

Regards,

AB.

Comments

  • This is my understanding - the OS (initiator) uses something called the Multipathing Algorithm to determine which fabric to use. It's not a load-balancing concept - they don't try to balance the workload over the fabrics. It simply chooses what it perceives to be the "best" path at any given time, and failing to the other fabric is an event determined by the algorithm.

    From the storage (target) perspective, it sends acknowledgements on the fabric it was received. It isn't using multipathing, so it doesn't look at cost.

    I'm admittedly an R&S guy but this is how I understand it. Maybe someone else could chime in and verify :)

Sign In or Register to comment.