we have a UCM Cluster in Europe (Germany) and we would like to add some Subscribers in USA but we have 125 ms RTT over MPLS WAN. In the Design Guide (SRND) for UCM 8.x there is a given number of 80ms RTT but I would like to ask if this is the absolut maximum or if it would be possible to have WAN Cluster with RTT about 125ms?
Thanks for hints and answers.
The documented guide lines are what is supported. While you can have a subscriber over the WAN with a RTT greater than what is documented, you may experience problems. The value of 80ms was documented as the maximum RTT for a reason.
I guess this is one of those questions that relates to what will work versus to what is supported.
I've clustered CUCM over WAN links with higher latency than 80ms (upwards of 130ms) and it seems to work ok. We sometimes see strange things with BLF on handsets (no surprise there) but other than that it seems to work ok. Checking the dbreplication on the Pub and Subs always seem to show that things are sychronised.
HOWEVER, I (and my customer) know that this environment wouldn't be supported by Cisco. If we have any issues with database replication then they would not support us, and would be perfectly correct not to do so. If you go against what is in the SRND guide this is the risk that you are taking.
You really want to stick to what's in the guide. I would guess that the latency you listed is typical (or minimum), and at some points may be much worse.
I wouldn't be overly concerned about database replication as such, more about SDL signalling - if one CUCM decides that the latency or packet loss to another node is too high, it will just mark that node out of service which effectively partitions your cluster temporarily. This can cause all manner of unpleasant side effects.
May look into a seperate cluster for the US; interop between clusters is getting much better (e.g. cross-cluster EM).
Please rate helpful posts..