Microphone, Camera, Action! IP Video Conferencing Meets The Desktop
By Fran�ois Desloges
IP-based video conferencing promises as many benefits as its predecessor, Voice
over IP (VoIP). However, both solutions rely on the deployment of a quality of service
(QoS) IP network to succeed. Solutions exist today that facilitate the upgrading of the
data network to add QoS capabilities, and unleash the full potential of IP video
For a few years now, different vendors have offered interoperable video conferencing
systems compliant with the ITU H.320 standard. These systems rely on the worldwide ISDN
network to offer good end-to-end video conferencing solutions comparable to
traditional voice services.
Since H.320 video conferencing systems hinge on the same circuit-switched network as
the legacy voice industry, they suffer from the same limitations that justified the
creation of VoIP and Internet telephony. The intrinsic inefficiencies of circuit-switched
voice networks generally cause services to be expensive, especially long-distance calls.
And the traditional model of the intelligent, circuit-switched network is limiting the
deployment of new applications as well.
A VALUE PROPOSITION
Video conferencing systems using the IP network, based on the ITU H.323 standard, offer
the same advantages as VoIP in terms of cost savings and applications. Cost savings are
made directly by eliminating circuit-switched long-distance fees. The merging of the two
distinct circuit-switched and packet-switched corporate networks in a single, simple, and
scalable IP network also promises significant savings in infrastructure and management
costs. The other benefit is the new applications that would be developed around the model
of an efficient "dumb" IP network, with very powerful and versatile computing
end nodes. Applications integrating voice, video, and the Web in a powerful collaboration
tool promise significant day-to-day productivity gains to work groups.
All the benefits of IP-based H.323 video conferencing come at a cost. Indeed, the
legacy packet-switched IP network is as inadequate to carry voice and video as a
circuit-switch voice network is at transporting data. IP networks must therefore be
upgraded to offer QoS to video streams. However, this entry cost may not be as high as it
seems at first. Contrary to circuit-switched equipment, packet-switched equipment follows
a tradition of simplicity, scalability, and bandwidth inexpensiveness that stems from the
necessity to follow a historic 1,000-percent annual bandwidth growth.
However, QoS enabling an IP network implies two new capabilities. First, the IP network
must be able to sustain guaranteed bandwidth through time to Real-Time Protocol
(RTP)-based video sessions that do not adapt to bandwidth bottlenecks the way legacy TCP
data sessions do. Indeed, for interactivity purposes, RTP sessions do not retransmit video
frames if theyre dropped in an IP network bottleneck they are lost forever.
Guaranteeing bandwidth to video sessions ensures that as few RTP frames as possible get
dropped, and that the quality of the video session will not suffer.
Second, the QoS IP network must offer minimal latency and jitter (latency variation) to
delay-sensitive RTP video streams. Again, for end-user interactivity purposes, video
streams should offer end-to-end transport latencies of less then 100 ms and as little
jitter as possible. This is a challenge, considering a latency of over 120 ms is
introduced by a single, traditional 10 Mbps Ethernet switch with 100 frames of 1.5 KB
waiting in queue. Low latency and jitter are normally achieved by forwarding the frames
belonging to an RTP video session in high priority. By and large, voice streams used in
Internet telephony and H.323 video streams require the same QoS from the IP network.
However, video sessions typically use much higher bandwidth than voice sessions. That
characteristic requires special care be given in certain areas of the IP network.
THE H.323 DEPLOYMENT AGENDA
Considering the benefits associated with IP-based H.323 video conferencing, it is very
likely that H.323 systems will be used instead of H.320 ISDN ones, everywhere IP networks
have become QoS-enabled. And as the IP infrastructure evolves over time, there is a clear
potential for H.323 to take over the world. Unsurprisingly, this will not happen
overnight, and each portion of the IP network will show a different upgrade agenda.
On the public Internet, H.323 will not really be usable until a critical mass of
Internet service providers (ISPs) have implemented QoS in their backbones. History will
tell if the traditional carriers will rush to convert the huge circuit-switched network
theyve invested heavily in, or if only new competitive carriers will build such QoS
Utilization of H.323 over the wide-area network (WAN) will be limited by the bandwidth
available from different access technologies for the last mile as well. Indeed, video
conferencing sessions, commonly using 128, 384, and 768 Kbps, are much more
bandwidth-hungry than typical voice sessions at 10 Kbps. Although legacy 56K modems and 64
Kbps ISDN access lines were enough for a mix of VoIP and data, broadband access services
(xDSL, cable modem, etc.) providing more than 1 Mbps of bandwidth will be essential to
deploy H.323 video conferencing on the WAN.
However, corporations can start benefiting right now from H.323 advantages by bringing
video conferencing out of the local conference room where it normally sits directly
on an H.320 ISDN connection to the desktops of everyday users.
QOS-ENABLING THE LANs
Typically, only a few groups of users may need
desktop video conferencing (DVC) at first, and chances are that the corporate backbone and
part of the networking infrastructure will not need to be QoS-enabled for a while.
A good immediate solution consists of:
� Regrouping the DVC users on a single LAN subnet (using DHCP for easy adds, moves,
� QoS-enabling that LAN for all DVC users.
� Using an H.323/H.320 gateway to provide the group with top quality point-to-point
WAN connections over ISDN, until the Internet upgrades to QoS capabilities.
New Ethernet equipment exists today that provides QoS to voice and video streams on the
LAN, and takes care of both the guaranteed bandwidth and the latency and jitter issue,
while minimizing the impact of video addition on legacy data applications.
GUARANTEEING BANDWITH TO VIDEO SESSIONS
It is important to understand how guaranteed bandwidth is essential to providing QoS to
video sessions, since a video stream can easily require 12 to 76 times more bandwidth than
an IP voice stream. Video session bandwidth usage can quickly sum up to a large part of
the total available bandwidth on a pipe.
Fortunately, on the LAN, intense competition among equipment manufacturers has reduced
the price of bandwidth for a while. This very low LAN bandwidth price can be leveraged to
offer guaranteed bandwidth through the creation of a full bandwidth hierarchy, where
typically no more than 10 full duplex links of 10 Mbps share a full-duplex
100 Mbps uplink in an Ethernet switch.
Although such bandwidth provisioning was overkill in the TCP-data-only era, it
isnt anymore with RTP video sessions that do not adapt to bottlenecks. Besides,
Ethernet bandwidth has become so inexpensive that it now makes business sense to create a
full bandwidth network on the LAN, since it avoids the costs of RTP bandwidth
micro-management. Indeed, with such a simple solution, there is no costly management of
bandwidth required, since the network topology guarantees full bandwidth to all desktops
on the LAN subnets.
MINIMIZING LATENCY AND JITTER
On a LAN where bandwidth bottlenecks have been eliminated, the issue of minimizing the
latency and jitter of video stream can be fixed today. Indeed, the only thing Ethernet
equipment must do is to retransmit in high priority the delay-sensitive packets belonging
to RTP video sessions in front of the legacy TCP data frames.
New QoS Ethernet equipment performs this type of prioritization, with automatic
queuing. Video prioritization is achieved by performing deep inspections on each frame in
order to identify if it belongs to an RTP video session. The RTP information that is used
for time-stamping purposes by voice or video sessions is encapsulated at the fifth ISO
layer inside an Ethernet frame
Although five-layer deep inspection involves huge processing power, this prioritization
mechanism requires no special management software or Ethernet adapter drivers to be
deployed, resulting in a very low upgrade cost. It is also totally transparent to any
other managed prioritization mechanism, and therefore, avoids being locked with a single
vendor solution. RTP auto-prioritization modularity and simplicity allow progressive
QoS-enabling of the network. There is no need to replace the entire existing
infrastructure, and users simply add QoS-Ethernet switches along the path between the
video conferencing desktops and the H.323/H.320 gateway.
It is also a good idea to use an H.323 gatekeeper on a QoS-enabled subnet to limit the
bandwidth that the video sessions will use on your LAN so that your legacy data
application will still have bandwidth space to move. Beginning to deploy QoS equipment on
LANs today makes business sense. Since it locally connects video conferencing desktops,
only the corporate backbone will need to be upgraded when enough ISPs offer the QoS
Internet access that will give its full value to DVC.
Fran�ois Desloges is director of technology for VIPswitch, Inc. of
Brossard (Montreal), Qu�bec, Canada. VIPswitch is the developer and marketer of
video and voice QoS Ethernet switches. The VIPswitch family of non-blocking switches
automatically prioritizes delay sensitive video and voice to the desktop with QoS and 100
percent throughput. The products are performance enhancers to all H.323 systems, and are
fully compatible with existing software, hardware, and networks from Cisco, Nortel, 3Com,
and Lucent. For more information, visit the companys Web site at www.vipswitch.com.