The first step is to put the telepresence users together with the telepresence vendors and figure out what resolution and what bandwidth is required for your application. There is a tradeoff here between visual quality, visual motion handling and the speed of the network. Some enterprises are happy with a lower bandwidth connection, while others require full speed. Be sure the important end users (read as 'executive staff') see and understand the quality differences before agreeing to a lower-than-full-speed bandwidth approach.
Next, do the calculations. Some telepresence vendors will specify the bandwidth as the transport bandwidth (the amount of video traffic that needs to be carried), and some will specify the actual IP bandwidth needed. As a network guy you want the latter. An HD codec running at 4 Mbps transport requires a bit less than 5 Mbps of IP bandwidth, so the difference can break your design.
Let's assume for the moment your enterprise has chosen a 3-screen system from one of the major vendors that you plan to run at full speed. Most of these systems today require about 5 Mbps' IP bandwidth per screen, so that means 15 Mbps per telepresence suite. If the WAN is an MPLS cloud, we need 15 Mbps of full-duplex bandwidth for video at each site where a telepresence suite is deployed.
Some of the telepresence vendors create multipoint calls by directly connecting telepresence suites to each other. This works for small numbers of locations. For instance, with the three-screen systems in our example, we can connect each site to three other sites by connecting one screen to each other site. Usually when the systems are connected this way, the center camera zooms back to allow it to take in a view of all the people positions in the room. So then this one image of the room is sent to one screen of each of the other systems. So in my telepresence room I get an image of the whole room of site A on my left screen, Site B on my center screen and site C on my right-hand screen. Thus, you can connect a total of four sites together without the use of a Multipoint Conferencing Unit (MCU). The bandwidth required for this 4-location call is no higher than if the endpoints were connected as two point-to-point calls since each endpoint still only requires 5 Mbps per screen for a total of 15 Mbps per location.
For larger endpoint configurations where an MCU is involved, it is important to understand the bandwidth requirement of the MCU (also called a 'bridge'). When a multipoint call using the MCU is active, each telepresence suite connects to the MCU. The MCU then formats the right video image for each screen of each telepresence suite and sends that information back to each site. So for our 3-screen suite example, there is a 15 Mbps bandwidth requirement between each suite and the MCU. The bandwidth demand at the MCU can add up to large numbers quickly. A 10-location telepresence call with 3-screen suites requires (15 Mbps X 10) 150 Mbps of bandwidth at the MCU. Placement of the MCU is critical because of its high bandwidth demand. One choice is to locate the MCU at corporate HQ or a data center where high bandwidth is available. A second choice is to host the MCU in a collocation facility with the service provider or a third party where access to the core WAN bandwidth is much less expensive.
Lastly, determine the bandwidth requirement to the telepresence management facility or concierge service. The management facility requires connectivity to each location and a certain amount of bandwidth to monitor the operation of the room. If the management facility also participates in the call, video bandwidth to the management facility is required.
The bandwidth requirements of the telepresence application will drive the operational cost of the network connectivity, so work through this design with care and determine the right tradeoffs for your particular implementation and needs.
