12-28-2011 01:58 PM - edited 03-14-2019 09:06 AM
We have two CUBE routers one in HQ and one in DR, calls are sent from carrier in round robin fashion. We have a CUCM cluster and UCCX cluster spanning the Metro-E.
HQ side-
Call Manager Subscriber (7.1.5.32900-2)
Contact Center Express Secondary (8.0.2.11004-12)
DR side-
Call Manager Publisher (7.1.5.32900-2)
Call Manager Subscriber (7.1.5.32900-2)
Contact Center Express Secondary (8.0.2.11004-12)
The DR CUBE router communicates across a 100Mbps Metro E to CUCM in HQ.
Normal operation is for CUCM to communicate with UCCX locally in HQ, uplinked with 6509 on Gigabit ports. The only traffic across the Metro-E would be the initial call from the DR CUBE to HQ CUCM.
This works normally, usually with no issues.
We recently had a problem with the HQ UCCX server and it failed over to DR UCCX, CUCM stayed in HQ. Under a light call volume, this worked with no problem, under heavy load, the processor in both CUBE routers, almost pegs, upward of 90% and higher, calls stopped processing.
It happens every time the UCCX server is in DR and the CUCM is in HQ, when the UCCX server gets moved back locally with the CUCM, the problem goes away.
The only thing I can see different is there is anywhere from 12-50ms of latency between the two servers across the EVPL, but if the DR CUBE is sending a call, then the path would be DR CUBE, to HQ CUCM, to DR UCCX.
Is there a potential problem with that much latency in the signalling path?
12-28-2011 10:38 PM
Well, the permitted latency in UCCX HAoWAN is max 80ms RTT i.e. max 40ms in one direction. If 50ms is seen in one direction, my first hit point would be that. Having said above, I am surprised with the latency reaching 50ms in an 100 meg pipe. May be this is something you should ask your data networking guy to look into ?
Also, check if your CUBE needs some IOS or resource ( memory etc ) upgrade.
Otherwise, I dont see any issue in the deployment & the functionality should work well.
Pls rate the post if it helps.
GP.
12-29-2011 05:20 AM
There is additional data traffic going across the EVPL, but the link is not getting congested. It does get upwards of 60-70M at times.
As far as the max latency within the UCCX cluster, would this also apply to the communication from CUCM to UCCX?
The CUBE routers have 1G of RAM, but if this were the issue, I would think we would see this problem all the time. The only time we have a problem is when the UCCX server is across the Metro-E from the CUCM server.
12-30-2011 11:48 AM
wilson_1234 wrote:
As far as the max latency within the UCCX cluster, would this also apply to the communication from CUCM to UCCX?
80ms RTT is between 2 nodes of UCCX but the value should not change even between UCCX to CUCM, because the other side UCCX & CUCM will be part of same LAN. So same latency figures applies to UCCX-CUCM across WAN.
HTH
GP.
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