I mentioned in my last post that I had a conversation with Jim Idelson of DesigNET, and we worked on two models of video conferencing utilization, one for room-based video and one for personal or desktop-based video. These two deployment scenarios have very different use characteristics, and so have to be considered independently.Room-based systems and Telepresence Room-based systems are used as part of traditional meetings. A meeting is scheduled, typically starts on the hour or half-hour, and has numerous participants. Room-based video conferencing meetings will often have multiple people in each video conferencing room. Calculating room utilization can be done by reviewing the schedule of calls that used the room and their duration, but knowing how many employees are using video conferencing requires also knowing how many people were involved in each video conference at each endpoint. This is typically more difficult to track.
According to Jim, typical utilization for room-based video conferencing systems is in the 10% to 25% range for well utilized rooms. The reason these numbers are relatively low has to do primarily with the common practice of using a conference room both for video conferencing and for local meetings. It is possible in a large facility to dedicate a room to video conferencing. But in smaller offices the video system often goes into the one conference room and so that room is by definition shared for all types of meetings.
Telepresence room utilization has been touted by the vendors as being at much higher levels. This has been backed up by some primary research by Andrew Davis at Wainhouse Research. This is true because of the high visibility of the telepresence investment and its use by executive staff, partly because it is a dedicated video room (it is rarely used for a local meeting or for a birthday party), and partly because it really does provide a superior remote meeting experience.
Personal Video Conferencing Personal video conferencing includes software on the desktop or laptop computer, and also includes dedicated executive appliances like the Tandberg 1700 MPX or the Polycom HDX 4000. In both cases the equipment or software is dedicated to the use of a single individual. No scheduling is required because the user always owns the resource (i.e. if I am in my office I can use it). The utilization model for personal video is much more like the models we use for telephone use.
In telephony we use Erlangs as a way to estimate the number of concurrent calls needed on a telephone trunk. This model makes assumptions about the number of employees who are simultaneously making phone calls, and the length of the average call. Although this model has served telephony well for many years, the changing technology may mean we cannot directly apply this approach to video.
The length of the average phone call may be increasing because many of the quick phone calls we used to do ('ready to go to lunch?') are now handled through instant messaging (IM). In the same way we probably don't step up to a video call unless we have a longer conversation in mind than what can be handled on a quick voice call. Meetings are still likely to start on the hour or half hour, although some ad-hoc video calling will also take place.
So we need a statistical model like Erlang, but with a different set of characteristics to determine the concurrency of calls and the length of calls so we can accurately estimate the peak bandwidth demand on the network. Note that utilization of an individual personal video endpoint is not as important from a financial perspective, especially for low-cost investments like software-based clients, but the combined concurrent demand of enterprise-wide desktop video on bandwidth is a very critical design point and needs to be estimated accurately and then measured to ensure the necessary resources are in place to support the video demand.
