Best Practice: MRG and MRGL with 3 servers and no hardware

[catkins101]

I am running CUCM 12.5 with three servers in the cluster. At some point in the past, a prior employee set up the systems with one single location in mind and had two servers. After some time, another location was added and a third server was introduced to the cluster. The device pool was never divided up so all devices are in one large pool and have same MRG. MRGLs and MRGs did not look optimal to me. So I ran a report on which conference bridges were in use to find that mostly CUCM was choosing the conference bridge resource on the server in the second location. This means any time someone does ad-hoc conferencing the traffic will come from the gateway in the first location and travel across the WAN to the second location where the third server is located. I could remove the resource but would like to leverage it. What is the best way to handle this? I was going to create MRGs based on server, but I read that the MRG should be based on resources. This causes a problem since MRG resources are not prioritized top down like the MRGL. It would seem easier to prioritize if I put all resources for server A in one MRG, all resources for server B in another, and so on. Then I could create an MRGL with the ability to put the MRGs in priority of which server to draw resources on in order. I am not using the hardware on my gateways at all right now.

[Roger Kallberg · Accepted solution]

You can still have it based upon resource type and split it up into different MRGs based on locality. Although I see no actual problem with having a mix of resources in a MRG if you want that.

[Maren Mahoney]

When a media resource registers with CUCM it tells CUCM about it capabilities and also its capacity. Two resources offering the same thing (conference bridges, for instance) that are in the same MRG will be load-balanced with CUCM allocating from the highest capacity resource first. This is true not just for two CUCM servers both running the IPVMSA service. For example, if you have a DSP-based conference bridge with 16 threads (maximum sessions 2 in the dsp profile) and a CUCM server with the standard 48 threads and both are in the same MRG the CUCM server will allocate from the server-software CFB first because it has more capacity and will attempt to allocate a transcoder if needed, even though the hardware CFB can handle multiple codecs.

I do want to ask: Is the IP Voice Media Streaming App service running on all three of your CUCM servers? If so, then it is no problem to put each CFB in its own MRG and order them in different MRGLs based on how you want CUCM to allocate resources.

Side note: If you have a single device pool for all of your devices, that means that all of your devices are using the same CallManager Group ("Cisco Unified CM Group"). If that is the case it means that all devices are registered to a single CUCM server (with failover) rather than being load-balanced. I can see a use-case for doing exactly that if that is your intention. But if you are looking to optimize your cluster (especially if you are going to be creating additional device pools for the conference resource issue) it might be a good idea to examine the CallManager Group configuration.

Maren

[catkins101]

Very helpful information. I do have IPVMSA service up on all three servers. It sounds like I may need to think more about the overall design and not just MRG and MRGL arrangement.