BladeSystem Virtual Connect
Showing results for 
Search instead for 
Did you mean: 

Question about failover latency on VC

Trusted Contributor

Question about failover latency on VC

Jun had a customer quesiton on how long it takes for VC to recognize a failed uplink and failover to a backup link:



    I know that LLDP on VC can find the loop in the network and automatically disable the uplink port on VC. If the active link failed, how much time will be taken to fulfill the failover? In general, for common network switch via spanning tree protocol, the link takes about 10s - 40s to recover, that is the point why my customer concern about the latency. 

    Many thanks for your feedback.   



Comments back from Paul:



Most VC failover testing I have performed shows a failover to take 3 to 4 pings, closer to 10 seconds than 40 seconds.

LLDP has very different role than STP, and it is not involved in loop detection/prevention, other than characterizing what is at the other end of the link for initial configuration. 


Link Layer Discovery Protocol (industry standard) is akin to Cisco CDP (Cisco Discovery Protocol), and is used to discover the capabilities  of the port at the other end of a link – chassis name, port #, VLANs, LACP status.  So, LLDP is what VC uses to figure out if multiple links in a Vnet or SUS can be active/active or must be active/standby.  From that aspect, it prevents loops at initialization time for a Vnet or SUS by making the appropriate active vs standby configuration on an uplink by uplink basis.  However, once initialized, I don’t think LLDP has any role in loop detection or prevention, just like CDP does not.  To summarize:


LLDP is industry standard equivalent of Cisco proprietary CDP

Unnamed proprietary anti-loop algorithm is HP equivalent of Cisco proprietary STP



Have any of you tested this in your environments? Let us know.