BE6k and BE7k - Design -CUCM

[Manish Prasad]

Hi Experts,

We have a customer running Cisco Collaboration Applications on BE6K servers, they have two servers one is Publisher and other one is Subscriber. As we know BE6K only supports up-to 1000 CUCM users , now the customer is increasing their users count to 3000 users by this year end. Their requirement is to add servers to support more users.

The solution that we can propose is to add two more BE6K servers to increase the capacity of the users to 3000 or one BE7k server to support the remaining 2000 users and more ( as BE7k supports upto 10k).

My question is that if we propose solution for BE7k server , do i need to change anything on the Publisher ? As my Publisher is running on BE6k server with 1000 users OVA template. Will that publisher (BE6k) work fine in DB replication for 3000 users ?

Note- All the phones are registered to Subscriber , other services like TFTP ,Extension Mobility is also running on Subscriber.

Thanks

Manish

[Atul Gaba]

Adding a BE6k subscriber will only provide redundancy and will not increase the user cap of 1000.

-Atul

[Suresh Hudda]

You need to migrate to BE7K bcoz BE6K only supports 1000 users, and you can't have BE6K and BE7K as part of one cluster.

Suresh

[Manish Prasad]

Thank you Atul and Suresh , so guys you mean that we have to replace existing both BE6k servers to support 3000 users ?

I mean if we add two more BE6k with 1000 users registered to each node , we can achieve 3000 count.

[Suresh Hudda]

Yes, it need to be replaced. As Atul said increasing sub will only increase redundancy.

Suresh

[Jonathan Schulenberg · Accepted solution]

I wanted to chime in with a bit more detail on this one since the cost implications aren't trivial.

Software

The cluster capacity of CUCM is limited to the performance [size] of the smallest/weakest cluster node. Remember that the cluster operates in a full mesh at the database and SDL layer. If the OVA size of the publisher is 1000 users then that is also the cluster capacity, even if you add a 7500 user OVA subscriber.

You can redeploy the cluster with DRS backup/restore to larger OVAs or use Prime Collaboration Deployment.

Also, be aware that whenever the cluster exceeds 1250 devices, the SRND calls for a minimum of five (5) nodes: publisher for database only, two TFTP nodes, and two call processing nodes. Ensure that you size the new cluster correctly by breaking these roles out.

Hardware

The BE6k and BE7k are little more than a marketing term created because customers want to buy appliances. It was done only to make the sales process easier: "here's our all-in-one appliance" and not have get bogged down in the technical design of a virtualization solution (e.g. compute/storage/etc). The Business Edition servers have the same hardware performance as the equivalent Tested Reference Configuration (TRC).

So, if you have the BE6k-Medium server, it is no different than the UCS C220 M3S TRC#2. Working within the sizing rules of that TRC, you may be able to rearrange the virtual machines and continue using the existing servers for some applications (though admittedly not CUCM or CUC); or, repurpose them for additional functionality such as Cisco Expressway.

The problem with the BE6k is the CPU which falls into the "Restricted UC Performance" category. You can review each application to see which OVA size is allowed by TRC or Business Edition model. For example, CUCM only supports the 1000 user OVA on the BE6k, not the 2500, 7500, or 10k OVA.

Suresh is correct that you'll need to replace the BE6k because of this if you wish to remain within the support model of the TRC, namely that Cisco guarantees performance of the entire stack. If you were willing to move into a Specs Based support model then you may be able to upgrade the CPU and other components* of the existing servers to support larger OVA sizes. This is where the design drifts into the long grass.

*Other components is most likely going to include memory and HDD storage due to IOPS. The BE6k has a limited number of spindles which will constrain the IOPS it can provide. You could either use SAN storage (e.g. FCoE, iSCSI, etc.); or, install Solid-State Disk to get the IOPS performance needed by the larger OVAs.

All of this is going to cost money, though probably a little less than replacing the servers outright. You have to decide which approach is better for this customer. Personally, I wouldn't venture off the TRC path unless the customer is really really good with virtualization and SAN storage. If they are though, it might be easier to just move it to their existing virtualization environment under the 3rd-party Specs Based support model. 

[santoshbansal17]

Hi All,

what is the best approach and/or methodology that we should follow to migrate From BE6K servers to BE7K servers. 

What would be the steps? do we need to have a downtime, is there a document that is available as a best practice.

Please help.

Santosh Bansal

[Jaime Valencia]

You can simply DRS from the lower capacity VMs from the 6K to the 7K machines, 6K and 7K are merely marketing terms, the versions of the products are all the same, just different capacity.

[santoshbansal17]

Sure thanks. 

I thought we need to do :

1. Assessment of the current servers

2. Backup the existing servers (BE6K)s

3. Construct BE7K Servers

4. Migrate from 6K to 7K ( 1 by 1) or lift and shft

5. Test 7K environment

6. Release to production environment.

[Jaime Valencia]

You're getting the wrong idea here, the limit of a 6K is PER cluster, NOT PER node.

If you already took the idea of adding more servers to a 6K into practice, because you thought that each node would add 1K users for the overall capacity, that means all of those deployments lost all Cisco support by doing that.

You can have 8 processing nodes + PUB on a 6K, but you still have the 1K users limit, and the 1250/2500 devices limit across the whole cluster.

The only option, is to go either the 7K route, or regular UC on UCS, or 3rd party HW if they already have servers with enough resources to have the apps.