More than 1000 users - Distributed call processing model - BE6k bundles

[Andriy Zatyshnyy]

Hi all!

I made distributed call processing model with BE6k bundles.

5 sites = 750+450+2x180+90 =1650 users.

399 VM = 180+108+2x43+25.

Will it fit at all?

P.S. 5 CUCM and 5 CUC clusters being on UCSs clusters BE6K-STBDL-PLS-K9= with VCenter +vSphere Ent plus.

KR

Andriy

[Aman Soi]

Hi ,

Whatever deployment model you use , u cannot have more than 1000 users in BE6K.

regds,

aman

[Andriy Zatyshnyy]

But i got 5 BE6k units each less than 1000 users.

KR

Andriy

[Aman Soi]

Hi,

I think u have got 5 piece of  Hardwares which would be deployed at 4 remote sites with one central site .Am I correct?

What is the H/W u have purchased?

This whole would form one single cluster supporting total of 1000 users.

regds,

aman

[Jaime Valencia]

Assuming you have 5 separate clusters, each one not going over 1K users, and not one cluster with 5 nodes, then yes.

You can have more than 2 servers in a 6K deployment, but if the user count goes above 1K in any given cluster, then it won't be considered a 6K but rather a regular CUCM cluster.

HTH

java

if this helps, please rate

www.cisco.com/go/pdihelpdesk

[Andriy Zatyshnyy]

Yes it is, exactamente Jaime!

these are my design considerations

5 sites=5 seperate clusters--each 2 nodes

each cluster no more than 1K users

each cluster =2 servers c220 M3S being in bundles+VMware

KR

Andriy

[Chris Deren]

You don't want 5 separate systems to manage, unless they are in different parts of the world for example. Ideally you want to fit everything on one solution, so rather than going with BE6K bundle implement CUCM/UCXN/etc which are exactly the same.

HTH,

Chris

[Andriy Zatyshnyy]

Chris all sites are in one city.

My design suggestions between distributed and centralized model.

Bundles are nice cost-effective products in this situation.

Will it fit centralized call processing model with BE6K boxes or no?

I study CIPT1, all solutions there before BE6K.

Got an experience in clustering CUCM and CUC in one site by classic scheme.

Thx in advance!

Kr

Andriy

[Chris Deren]

This is a very bad idea, remember you lose features across clusters and management becomes difficult. Think about future upgrades, dialing between sites (your design will require full mesh of SIP trunks), etc, etc.  What about if you want to add applications such as contact center express, attendant console they can only be integrated to a single cluster so now you need to buy multiples of those.  Deploy single traditional CUCM cluster with traditional Unity Connection to support 1600 users, cost wise you will not save much with BE6K licenses and in fact you will need to buy more hardware (5 servers) where you can nicely fit everything on 2 UCS servers.

HTH,

Chris

[Andriy Zatyshnyy]

Thx Chris!

now i see how it had better to choose.

Let it be classic CUCM+CUC.

Multisite WAN with Centralized call processing model with single cluster in HQ:

2 UCS c220 M3S

CUCM 9.x cluster:

CUCM Pub+TFTP on one VMWare instance at 1st server

CUCM Sub or Sub1+Sub2 on other VMware instance at 2nd UCS

CUC 9.x cluster:

CUC Pub on one VMWare instance at 1st server

CUC Sub on other VMware instance at 2nd UCS

I guess standard vSphere lics will be enough per CPU or i should take vCenter with vSphere Ent Pluse lics (no cost effective).

Guess ISP gave one private network over fiber. Now they have 50 Mbps each site WAN connection.

Should i propose to buy Cisco IOS Gatekeeper options for all the sites?

only G.729 over WAN for all sites?

KR

Andriy

[Chris Deren]

I would deploy as following:

server 1: CUCM Pub(ELM), CUCM Sub2, CUC Sub

server 2: CUCM Sub1, CUC Pub

Where the CUCM Group would be configured to use CUCM 1, CUCM 2, CUCM Pub

Assuming both servers are at the same data center.

No need for GK in your deployment, simply deploy voice gateways at the remote locations registered to CUCM.

In fact you should never try to deploy GK in any new deployments as GK is for H323 deployments and recommended protocol is SIP, so SME would be a replacement, but it is only desired (never required) in large multi-cluster complex environments.

I recommand you read Cisco CUCM SRND where a lot of this is well described.

HTH, please rate all useful posts!

Chris

[Aman Soi]

Hi Andriy,

You can refer the BE6000 FAQ for more details.(+5) to Jaime and Chris for explainations.

http://www.cisco.com/en/US/prod/collateral/voicesw/ps6788/vcallcon/ps11369/qa_c67-703226.html

regds,

aman

[Andriy Zatyshnyy]

Hi Chris!

Trying to dive deeper with my design (reading CIPT1 guide+SRND 9).

For now i can suggest that with my centralized model:

• all the phones from 4 remote locations will register normaly via WAN with ISP optical lnks (at least subnetworks should route each other) with SIP help and no GK, no SIP trunks etc.;
• i should be configuring MGCP gateways for PSTN (E1) links to backhaul all ISDN traffic to CUCM (SRST scenario).

Am i correct in my assumptions?

KR

Andriy

[Chris Deren]

Phones at remote sites will simply register with central CUCM over the WAN, no need for anything else here.

For GW control you can either use MGCP/H323 or SIP, I do not recommend MGCP (my and other's preference) as I have seen too many unexplained problems with it, my standard for GW control is SIP, so the GW is configured as SIP trunk in CUCM and proper dial-peers pointing to CUCM and TDM trunks.

HTH,

Chris

[Andriy Zatyshnyy]

Aha, thx Chris, i see my weak place=GW configuring with SIP.

i should read more about SIP trunks and TDM.

So it will be 1 GW in a HQ site and 4 in Brs.

in SIP Trunk Config being at CUCM 8.6 Lab demo i see PSTN Access checking box and SIP address of a Remote site...

Under Gateway section being at Cisco 2921 i see MGCP and SCCP choice...

KR

Andriy