Cisco IT currently has approximately 21 Cisco Unified Communications Manager (UCM) clusters worldwide to support our global voice communications in the production environment. Where we deploy these clusters around the globe reflects our user concentrations, their requirements for call performance, and the availability of and regulatory considerations for network infrastructure in different parts of the world.
We use Cisco UCM servers to support our corporate voice needs, and to support our customer contact centers around the world. For our corporate voice environment we have three different types of clusters:
- Campus cluster for our larger campus sites
- Regional Centralized Call Processing clusters that aggregate and control the traffic from many smaller Cisco sites
- Cluster-over-the-WAN clusters that support regional offices with servers based in different geographical locations
The campus clusters serve our campus locations in Irvine, California, in Research Triangle Park (RTP), North Carolina, as well as our headquarters location of San Jose California where we have the largest numbers of Cisco employees and phones. With the campus cluster, all call processing resources are local, so we don't need to provide resiliency in the form of Survivable Remote Site Telephony. In San Jose alone, the campus cluster can support up to 60,000 phones across our current 50 or more buildings. This cluster is our largest, and is considered a “super-cluster”. We support a Publisher, two TFTP servers (primary and backup), and 16 subscribers (half primary and half secondary), for a cluster size of 19 servers. We also support a couple of virtual server instances to provide IP Phone Services, as well as servers for voicemail (Unity Connection on UCS), Presence, and dedicated Trace File storage servers. We also have centralized some of our Cisco Unified Operations Manager and other management and provisioning servers there.
Our Centralized Call Processing clusters are located in each region to serve multiple Cisco offices while optimizing network performance, controlling costs and supporting central administration. The US East and US West clusters serve Cisco users in those regions including Canada, Mexico and Central and South America.
We always try to locate cluster servers in two different data centers in the same campus when we can, for added redundancy. In Cluster-over-the-WAN, we keep some of the cluster servers in another location entirely. Our Europe/Middle East/Africa regional cluster is one example of a Cluster-over-the-WAN environment – in this case the publisher and some subscribers are in Amsterdam, while subscriber and other servers in the same cluster are in London and Brussels; the distance adds redundancy without affecting service performance.
We have used these three models to support our standard environment. In the case of India, a Centralized Call Processing cluster is used to support the region. Likewise in the Asia Pacific rim we have several Centralized Call Processing clusters to support the many Cisco field sales offices in this part of the world.
Choosing locations for Cisco UCM clusters is like fitting together the pieces of a changing puzzle. Although the decision factors are complex, our ongoing goal is to consolidate the number of clusters to to support our global environment.