Fit Video to the Bandwidth, or the Bandwidth to the Video?
Constraining bandwidth may slow video conferencing uptake, but in some regions, bandwidth may be just too expensive. How to strike the right balance?
One of the key elements of my network assessments is to determine the bandwidth requirements for a planned video conferencing deployment. A few years ago this was a straightforward proposition just requiring a spreadsheet to add up the numbers. Video conferencing primarily occurred from conference rooms, and we could expect those conference rooms to be used at 100% utilization during the busy hours of the enterprise day or week. So the video conferencing demand was merely the number of video conferencing systems times the bandwidth required for each unit.
Today many of my customers are deploying desktop video conferencing in addition to room-based videoconferencing. Desktop systems will likely not be used 100% during the enterprise busy hour. Their utilization may look a lot more like telephones where we use Erlangs to determine concurrent utilization. But I don't think we can just directly use Erlang calculations for a number of reasons.
First, desktop video conferencing calls may often be longer than the average telephone call. Video is not likely to be used as a direct substitute for the telephone but instead used for meetings where it's important to get visual information and where a longer conversation is likely to occur. These conversations may not be as long as the one hour to 1 1/2 hour meeting we expect from conference rooms, however we likely won't use video to check on whether we still have an appointment for lunch or if that report is ready. It will be a meeting with some substance.
Secondly video conferencing calls will not connect back to a central IP-PBX, but will instead connect through the IP network directly to the other party involved in the call. Centralized connecting will occur for multipoint calls--calls that include more than two parties--but the majority are likely to be point-to-point.
My spreadsheets today have a couple of different approaches to calculating this simultaneous use factor for desktop video calls. One component is to try to guess how many simultaneous users occur in a small or large office. Clearly the statistics of small numbers apply. If an office has one or two desktop video conferencing systems, the probability of 100% concurrent utilization is very high. Conversely if an office has 200 desktop video conferencing units, the probability of 100% utilization is much lower and perhaps a number like 20% will suffice.
A second component of the spreadsheet tries to guess the concurrent utilization from all offices back to the central bridge resource. This concurrency may be affected by time zones, the uptake of video within a company (which may differ by region), and by the basic culture of the company. Some companies are very centrally driven, so remote offices are always calling back to headquarters. Some companies have more distributed control so there may be a higher volume of calling within the region and less calling back to headquarters.
Once we make our best guess, the question becomes does the company support the predicted bandwidth, or does the company deploy call admission control to limit video conferencing calls to the currently available bandwidth? This is partly driven by funding availability and also by the view of the company as to the value of video conferencing. Funding sources may add complexity because the cost of bandwidth may be borne by the IT organization and not by the businesses who use video conferencing. IT is traditionally a cost center, and is judged on reducing cost on an annual basis. Deployment of bandwidth to support video conferencing runs contrary to this trend. Conversely if businesses can increase productivity through videoconferencing use, it may make sense for the company to spend this money. Determining how to make this financial connection across the organizations of the enterprise is often difficult.
Some companies decide to limit video conferencing to the available bandwidth and wait for the users to complain. Other companies decide to provide sufficient bandwidth based on the spreadsheet analysis, and then monitor utilization to see if actual uptake matches expectations. This latter approach is more like the way a service provider runs their business, watching utilization and ensuring resources stay ahead of current customer demand.
Which is the right approach? I think the answer is specific to each company. Constraining bandwidth may cause video conferencing uptake to be slow. If the company is depending on video conferencing to increase productivity, this may be a mistake. Conversely, some companies operate in geographic locations where supplying the needed bandwidth may be just too expensive, and will thus need to limit use to the available bandwidth or downgrade the bandwidth used per call (as well as user expectations of quality.)