Hope everyone is doing well and thanks in advance for any inputs/suggestions !!!
I am working on upgrading my Unity environment from Unity 4.0.5 to Unity Connection 8.x. I have failover configured in my existing Unity 4.0.5 environment with Primary and Secondary server in separate building. I understand the architecture for Unity Connection is completely different from Unity 4.x where both servers are active and one is configured as a Publisher and the other is configured as a Subscriber kinda like CUCM. In my scenario I will have the Pub at the data center where the majority of the users are and the Sub will be at the DR site for backup only. I am planning on configuring the Pub to be the primary call processor and wonder if you guys see any issues/concerns there. I understand for best practice you want the Sub to be the primary call processor but in my scenario the majority of the users reside at the same location where Pub is located and the Sub is at a DR site connecting across a WAN link. If I have to make the Sub to be the primary call processor then the majority of the users will have go across the WAN to get their voicemail and that I don't that's a good idea. Let me know what you all think !!!
I can definitely see where you might run into some issues with this. The main thing is trying to balance out the processing load placed on the Publisher.
So, you have to think about the following:
1) The Publisher should be the primary interface for web applications - Administration (CU Admin/Service/DRF/Platform/etc) and User facing features (PCA/Unity Inbox).
2) Are you going to do integrated messaging via IMAP? If so the Publisher should be the primary interface there as well. Are your clients capable of using IMAP Idle (e.g., Exchange) or IMAP (e.g., CUPC). The protocol used takes up more resources.
3) The Publisher is responsible for replication and maintaining the DB.
So, if you think in terms of what is best practice vs. what you are considering - the issue becomes the additive load that is placed on the Pub. Not only does it have to do all of the 3 primary tasks above but now it will have to be the primary call processing agent. Depending on the size of your environment, the hardware used, the clients connecting to CUC, and the applications being used - you could be setting yourself up for some issues out of the gate. That's not to say that your idea can't or won't work but it definitely goes against what is documented and known to work well. Not everyone purchases HA so obviously in those scenarios the Publisher or First Node handles all of the traffic - so it can be done.
IMO, the best thing for you in your scenario is to swap the physical placement of the servers. Assuming you are within the specs for clustering between buildings with CUC, then I would place the Subscriber where your users are and put the Publisher in the alternate location. Alternatively, you could consider a scenario where you put both servers in one data center and then have a cold standby in your DR site. There is some work involved with getting the cold standby online but it is another viable option to consider.
Please rate helpful posts!
It's always difficult to recommend anything other than what is marked as best practice. Mainly because Best Practice can mean fewer issues in the future. That being said it is possible to configure the system for load balancing as the Design guide says as you said in you opening question.
- Balancing the Load of Calls That the Cisco Unity Connection Servers Handle
Although it is possible to balance the load of calls that the Cisco Unity Connection servers handle in a Connection cluster, we recommend that most call traffic be directed to the subscriber server.
You could try pushing calls mainly to subscriber and see what happens.
In the end if you had an issue TAC is going to suggest Best Practice if issues are seen.
You could possibly consider switching the location of the publisher. There could be reasons this may not be possible for your environment.
Also, you could consult your Cisco Account team to see if they have any customers that might be doing this differently than
the suggested Best Practice. They could also have other designs that could be considered.
But, do be sure to follow bandwidth requirements and any requirements in the System Requirements guide: