Intracluster Run-Time traffic

[annayuzhao]

In Cisco CCM design guide for a distributed single cluster call processing solution, it mentions that if BHCA is 10000, it needs 900KBPS for the run-time data replication between the CCMs across the WAN.

What I am not clear is when active ccm handles the call process locally why it has run-time traffic between the active server and publisher? I understand there is run-time traffic between the ccms when phone register with the ccm, but not quite know how run-time traffic involved when there are attempt calls. Thanks

[pwalenta]

There are a number of things that occur. You have CDR data being transferred. Phone state information is constantly synced. You also have database replication and directory replication as well.

Cisco's spec says that at 10000 bhca, it'll use about 900k of dedicated bandwidth to keep the sync of databases proper.

This is an absolute and I don't recommend going around this. Bad things happen when CM database sync's break.

[annayuzhao]

Much clear now.

But I still have the following questions:

1.Does the run-time traffic also store in the SQL database, such as phone state information?

2.The moment I make a call, is there any specificly call set up related traffic needs to be replicated?

3. If I don't have 10000 BHCA, what's the bandwidth required?Can I reserve a small amount of bandwidth for the replication? Thanks

[pwalenta]

1. Run-time traffic includes much of what is stored in SQL. THe fewer the changes, the less that gets replicated. Phone state information might be in there, I'm not entirely sure - but phone state information is passed between all servers.

2. The call setup does not involve any server other than the one the phone is homed to, until a phone state push or query is done.

3. The minimum *required* bandwidth is 900k. Cisco will not support doing a cluster over anything less than this. I have been down this road many times. It gets messy. Consider 900k your bottom limit.