I am looking into deploying Unity Connection 7.x with a Pub and a Sub connecting across a WAN link. I've gone through Cisco design doc regarding Unity Connection cluster requirements and the requirement is rediculous. Cisco wants a 50MB link with a round-trip of 20ms or less for a 48-voiceport deployment which is what I have. I am wondering if anyone has done any Unity Connection deployment over a 45MB DS3 connection with a round-trip of under 40ms successfully !!! Does anyone know the reasons with the strict requirements ? Thanks in advance !!! I appreciate any inputs/suggestions !!!
There is a cold standby model for Unity Connection. Essentially, you build the system and keep it updated to the same revision of code as the production server(s). However, you need to make sure you have current DRS backups. In the event of a failure, you could restore the DRS backups to the failover Unity Connection server (cold standby). This requires some licensing work and of course, good backups...but that's an option. The architecture for Unity and Unity Connection is completely different - in obvious ways - but also from a clustering perspective. There is no such concept in Unity.
Please rate helpful posts!
The reasons for the requirements are likely tied to a few things. My thoughts:
With Connection clustering, each server plays a specific role. The Publisher should primarily handle database, web, and IMAP connections and be a secondary for call processing. The Subscriber should primarily only take calls. So, you would be essentially piping all of your voicemail calls across the WAN for at least one or more sites - so you would need to take this call volume into account and you need to account for the CODEC being used as well as QoS there. If you misconfigure the ports and failover for line groups, the HA redundancy does not function properly so keeping calls to a site may cause more problems than anything else.
Replication. Due to the HA nature of the Connection cluster, everything is replicated almost immediately between servers. If you login to the Pub and make a change, it is replicated to the Sub. If you make a change in the Sub, it is replicated to the Pub. So, again - these are transactions that would start to take away bandwidth depending on how active you are with changes - not just administratively, but letting end users handle changes from within PCA as well.
Digital Networking. Connection doesn't really have a clustering over the WAN model. Instead, you'll note that it's distinctively referred to as deploying servers within different buildings. Essentially, Connection is designed to be a localized solution where you can deploy across buildings or possibly high-speed MAN but not really the typical WAN, persay. However, you can configure digital networking for up to 10 nodes (7x). When you do this, each site has a local VM system but you can network them together to "appear" almost like one large cluster. When you do this, SMTP is used for networking between clusters.
So, here's the bottom line - if you want support from Cisco, i.e., you want to be able to call TAC then you need to fit within the parameters given. Does that mean that deploying across the WAN won't work? Not necessarily but it's not supported and typically for good reason. Call volume, replication, and web operations could hog up more bandwidth than you expect and suddenly no one is happy.
Please rate helpful posts!