Unity Connection Fault tolerance within and between DataCenters

[asherif82]

Hello Guys,

I have that scenario that customer has two seperate data centers located in two different location and need to deploy unity connection redundancy within each data center and between data centers.Let's say that users in DC1 are utilizing CUC located in DC1 and users in DC2 are utilizing CUC located in DC2 .If for some reason DC1 is down all user should be able to utilize unity connection located in Datacenter 2 and vice versa. As far as I know unity connection cluster can handle only two servers (publisher and subscriber) .        

How this can be achieved in unity connection ?

Please advise.         

Regards,

Amr Sherif

[David Hailey · Accepted solution]

Unity Connection has an active/active configuration so the failover within the application is handled on its own.  What you are referring to is a the redundancy at the call handling / routing level.  For this to work, you'd have the following configuration in CUCM:

Ports configured for the Unity Connection publisher and Subscriber

2 line groups - one for the Publisher answering the ports, the other for the Subscriber - ex:  CUC-Pub_LG and CUC-Sub_LG

2 hunt lists - one prioritizes the Publisher ports with rollover to Sub, and the other for vice versa - ex:

CUC-Pub_HL contains CUC-Pub_LG then CUC-Sub_LG

CUC-Sub_HL contains CUC-Sub_LG then CUC-Sub_LG

Then you would have 2 pilot numbers and 2 VM profiles to direct calls to the server designated for each location.

So - you can do this above and it works.

However, Cisco recommends that you direct all calls first to the Subscriber with failover to the Publisher.  This still works; however, it may or may not make sense for you depending on where your users are in terms of location relative to each DC.

Hailey

Please rate helpful posts!

[Chris Deren]

David is spot on (+5), I have done this many times with "load balancing" the traffic between 2 CUC servers.  One addition I will make is that if you are using SIP integration rather than SCCP, simply create 2 separate RL/RG with reverse order of SIP trunks.

HTH,

Chris

[asherif82]

Hi Chris,

Thanks for the reply .it really reflect what the customer need between two data centers but only one thing is left which is redundancy within the data center itself such as in DC1 I need to host two servers of unity connection so failover would be happened within DC1 and if both servers are down should DC2 other two Servers be active. I know that the cluster is mainly up to 2 servers (Pub & Sub). Is there a way to make 2 pair servers in each data center for failover internally and if both are down then failover should be maintain by the other data center?

Regards,

Amr Sherif

[Chris Deren]

No, UCXN cluster is 2 servers, period.  You can have more clusters integrated to each other via intrasite link, but each cluster hosts their own voicemail boxes, etc.

HTH, please rate all  useful posts!

Chris

[David Hailey · Accepted solution]

Amr,

What you need to relay to the customer is that the architecture of Unity Connection provides redundancy using an Active/Active 2 server cluster that consists of 1 Pub and 1 Sub.  They can reside within the same datacenter or across 2 datacenters.  You can't expand the cluster by adding more nodes or anything along those lines.  You could have 2 separate Unity Connection clusters - 1 cluster in each datacenter - however, this still doesn't fit what you are describing. As Chris mentioned above (+5), the 2 clusters can be "networked" together but this doesn't provide "redundancy" where phones can hit either cluster and still receive the same service.  Quite honestly, most customers in my experience do not fall into a size or scenario where networking of clusters really becomes a requirement anyway.

The likelihood of both servers being down within a datacenter simultaneously is greatly reduced if you have physical redundancy ... i.e. you are virtualizing UC nodes across different physical hosts and not all on one.  If you are doing that then you still have a single point of failure which is contrary to the nature of the requirement for multi-layer redundancy. If the WAN link between the 2 DC can support a cluster, that's what I would likely recommend.

Hailey

Please rate helpful posts!

[asherif82]

Thanks guys that was very helpful.