Today, there are about 400,000 desktop video conferencing units populating enterprises and small to midsized businesses globally, and that number is projected to surpass 1.5 million units by 2013, according to Robert F. Mason, a principal research analyst at Gartner Inc.
Desktop video conferencing requires extensive network planning because of the ad hoc nature of the technology's use. With room-based video-conferencing systems, users have to book time to use the equipment, and there's an obvious limit to the number of video streams going out over the wide area network (WAN). Desktop video conferencing has no such limitations on use. Employees can open up a session on the fly.
"[Desktop video conferencing ] is more challenging over a wide area network because the capacity is, of course, on a unit basis, much more expensive, so it's difficult to just throw capacity at the problem," said Mason, who recently authored the research note
Desktop video conferencing network problems will kill user uptake
Network architects and engineers tend to "absolutely overestimate the attach rate of best-effort video conferencing," Mason said. High-level executives who have been spoiled by high-def video systems will be unlikely to tolerate a significantly degraded experience, he said.
"[Users] have expectations not only for ease of use but for quality and repeatability," he said. "But once that [desktop video conferencing experience] falls below the threshold … the challenge is not to over-engineer, but at the same time, to deliver a repeatable, sufficient-quality service."
'Ad-hoc' desktop video conferencing can choke WAN
Unlike traditional video-conferencing systems -- which require users to book a room, giving it a fairly consistent usage cycle -- desktop video conferencing can be impulsive. Users may start a session even for "novelty" purposes, Mason said, and network engineers who don't do the right logistical and technical planning may see fluctuating and unpredictable numbers of simultaneous calls at any time.
Desktop video conferencing also has no predictable duration, making it difficult to gauge network impact, Mason said. But on average, organizations can expect concurrent calls from one in 10 desks for active users to one in 40 desks for infrequent users.
Although one in 40 may not sound like much -- even for organizations with minimal adoption -- it can overload an already saturated WAN link to a branch office, which may have the capacity for only one or two concurrent desktop video-conference calls, Mason said. Networking teams should allot 300-400 Kbps for an average session, he said.
"Coping with ad-hoc demand [requires] a combination of things -- it's looking for a solution for desktop video that has gatekeeper and admission control capabilities," he said. "You can control it by who gets the client distributed to them, and you can control it to some extent by policy."
Desktop video-conferencing network planning: Logistical and technical
Meanwhile, vendors are beginning to realize that these key features -- policy management, admission control and gatekeeping -- need to be built into desktop video-conferencing systems, Mason said. But enterprises with older systems may still have to purchase proxy servers or gatekeeper devices separately.
Branch offices that rely on the public Internet -- versus an MPLS network -- will do well with desktop video-conferencing solutions that use the Scalable Video Coding (SVC) protocol, Mason said.
From a logistics perspective, he added, deployment should happen in "layers." Start first with the users who will spark the highest demands, in terms of quality and usage, and provision gradually down to more infrequent and fault-tolerant users. Otherwise, "novelty use" may take off.
From the engineering perspective, networking professionals can use quality of service to improve performance across the WAN, as well as call admission control to ensure from the start that the session will do well, Mason said.
But walk a fine line, he cautioned. Too much engineering to constrain desktop video conferencing usage may undo its objective -- enabling communication.
"You have to think about it like telephony. There isn't really a busy signal. The people who use it expect it to work," Mason said. "Think of it [as in] many ways analogous to SIP trunking. You've [allotted] this much bandwidth and [planned for] this many concurrent calls…. It's not a blocking network."
Let us know what you think about the story; email: Jessica Scarpati, News Writer