Callmanager Network Design Question

[vocollect]

We had a reseller implement a VoIP for our network at 2 sites (US and UK offices). Each site has 2 Call Managers (ver 3.3), one publisher and one subscriber. From a design view this is great for failover, but the admistration is complicated. We have 2 sets of directories to maintain and create. Should the design have one publisher and multiple subscribers to maintain a single directory. What is the 'rule of thumb' or best practice for setting up additional sites. We have another site opening in the US and I don't want to have to maintain another directory. Thanks for any help.

[darryl.wilson]

Hello,

This seems to be OK. How many users are each each site? Typically, when dealing with a WAN with that large of a georgraphic distance, splitting a cluster across the WAN is not recommended (You typically need about 1.5 Mbps of Bandwidth between CallManagers for SQL and less than 30 ms Round Trip Delay). Thus, a distributed archiecture as you describe. You may be able to implement a Centralized Design with SRST in the other sites (again depends on size of sites). WHat is your voicemail deployment? Each site has it's own voicemail server? From what I see, think it was probably the best approach but if these other site are very small (under 100 users), a centralized CallManager with SRST may be possible....Again, the voicemail strategy needs to be understood further...

Regards,

D

[vocollect]

We have 150 users at the US Corp site and 10 users at the UK remote site. The WAN link between the sites is 1.5Mb point2point T1 with 100ms latency. The voice mail is Unity Express, one local at each site.

I was told that I needed 10Mb link between Publisher and Subscriber, is this true? Looking for a second opinion.

[mserwon]

The voice mail is unity, not unity express, at the US and UK sites.

[vocollect]

Thanks Mike for the correction. I got the IPCC and Unity version confused.

Have you come across anyone that is running a subscriber across the Alantic with latency 100ms+. I am hoping someone out there has the same concerns we have.

[marcusl]

Hi,

IMHO it's overkill to use a separate cluster for 10 users. This deployment should be able to use centralized call processing, however, it would require a termination point for the PSTN in the UK, but this would not be a problem to configure. Of course - you would have to consider load on the T1 aswell, but as long as that is not an issue, the centralized processing should be preferred.

Best regards,

/M

[smburke]

Having experienced the nightmare of the centralized call processing deployment over long distances in the US, my opinion is that it's not overkill to have a separate cluster for the UK office. Considering the stability and latency of the transatlantic links, combined with a radically different numbering scheme, different gateway format, etc. I know I'd rather have a dedicated CM at the UK site. If there are only 10 users, the move/add/change count can't be all that high, so aside from the initial investment and software/hardware support costs, it shouldn't take much to maintain the smaller site.