by Jim Frey of Enterprise Management Associates
A lot of attention has been paid to the role that video is playing in driving growth in overall network bandwidth consumption, but most of that is trained on the well-known, easy-to-name, sexy types of video traffic like YouTube, Skype, and telepresence. But woe is the network manager who ignores what is happening in the world of digital video surveillance, which can also generate big volumes of streaming traffic across shared network infrastructures.
In most shops, video traffic on the network is comprised of a number of different contributor types, each of which has its own unique set of bandwidth requirements and delivery constraints. Let's look at a few that are most talked about these days:
Multimedia Web content
Most of this is a combination of Web-based sources offering video content on-demand; the most well-known being YouTube. While this can eat up a fair bit of bandwidth (particularly if something is going viral) it is not typically considered real-time. Most YouTube videos are cached and invoked on-demand. As a result, while multimedia video can be annoying it can be accorded “best efforts” priority for network delivery.
Live streaming video
This is similar in many ways to multimedia content, because the traffic only goes one direction. The use case is one where a video stream is broadcast from one source to many viewers: think quarterly internal company meetings or March Madness basketball games.
Since the desire is to do this in a real-time mode, it is necessary to ensure that sufficient bandwidth has been reserved for the event, lest the network be too congested to deliver adequate quality. March Madness streams can gobble up 100-200kbps each, but high-end corporate video broadcasting can quickly cross the 1Mbps threshold. It’s also important to take advantage of network features such as multicast in order to de-duplicate streams as much as possible. There are other de-duplication options available, such features offered by WAN optimization controller vendors Blue Coat and Riverbed.
Desktop video conferencing
The poster child for this type of traffic is Skype, which is an adaptive program that can operate using as little as 30 Kbps but which will consume as much as 220 Kbps if available. The end-user experience is not great, but the expectations are also low. While many enterprise network managers simply wish to block Skype, it has become common tool for both voice and video communications and now that Microsoft owns it, its future looks bright.
There are many other solutions that fit this category, such as Microsoft’s Lync unified communications solution, which delivers low to moderate resolution but true real-time interactive video. This requires tight tolerances for latency, jitter, and packet loss along the way. The sheer volume of endpoints potentially driving such traffic could easily result in a significant footprint on the network.
As a real-life example, Avistar Communications, which focuses primarily on this type of technology, has a successful deployment exceeding 10,000 desktop video conferencing endpoints. Even at highly compressed, 384 Kpbs default rates, even moderate concurrent use results in numbers that rapidly reach frightening levels. The network manager in charge of that huge deployment has mandated a 30 percent bandwidth queue dedicated to low latency traffic to all sites, with at least half of that set aside solely for videoconferencing.
The Kingpin of live video communications, telepresence systems leverage multi-channel high-resolution video and audio to achieve a functional alternative to physical presence: site and sound, though (thankfully) not smell, at least for now. Consequently, it also has the greatest requirements for bandwidth and network delivery priority.
Room-based telepresence systems require 4 to 5 Mbps per CODEC for startup, and a sustained 500 Kbps to 1 Mbps per CODEC. The largest deployments of systems in the world today number over 1,000 endpoints per organization, but even a single pair of endpoints require network capacity planning, policy configuration, and sustained monitoring to assure session quality. (For more on those latter two points, including the case studies mentioned, see EMA’s Oct. 2011 research on Videoconferencing Impact on Network Management)