Not many people know that H.323, the umbrella under which the ITU (International Telecommunication Union) defined the protocols for the transmission of unified communication over packet networks (back when it was simply called “audio-visual communication”), was conceived in the good old ISDN days.
Nowadays, H.323 is mainly used for video conferencing over IP networks, which introduces new challenges that were not present before.. One major challenge is sending media that requires high bandwidth (especially since the mighty HD became a reality) over a packet switch network with variable bandwidth, variable packet delay (jitter) and packet loss. Another challenge is the computational complexity of encoding and decoding the transmitted bit stream. The more complex the video standard, the more energy (computational power, time, memory) it requires from both sides.
These challenges have been occupying minds in the visual communication market for a long time and special mechanisms have been developed to try and deal with these problems. On the one hand, infrastructure is frequently becoming better, both in terms of bandwidth and quality of service (QoS) and also in terms of the end-equipment that is used in unified communication deployments (particularly in large enterprises). On the other hand, the demand for this infrastructure is rapidly growing at a very fast pace and therefore bandwidth available for unified communications is still a serious problem. The endpoints that were bought last year already struggle to support the current capabilities of new endpoints.
This infrastructure problem is similar to that of an assembly line, similar to the one depicted by Charlie Chaplin in “Modern Times“. On the one side, the feeding machine which is operated by a technician, feeds the assembly line with pieces of machinery at an unknown pace. On the other side, the weak assembly line worker, is forced to adapt to the pace of the machine without any disregard to the pace of the machine and/or the assembly line.
The assembly line scene in Modern Times by Charlie Chaplin.
Even if you are not a certified industrial engineer, you can probably solve the assembly line problem. All you need to do is to allow the assembly line worker to control the speed of the assembly line and the feeding machine. This can be done either by giving the worker control over the assembly line and machine, or by regulating some sort of feedback from the worker to those responsible for the assembly line and machine (for instance, a good old “Slow down, god damn!”).
It’s true that one can leave the control of the assembly line and feeding machine to others (technicians, for instance), who would try to adapt the pace to that of the worker. However it is clear that in both cases – the assembly line speed (or bandwidth problem) and the feeding machine speed (processing power problem) – the receiving side (the worker) is the one who can make the best decision.
In video conferencing over IP networks, the bandwidth problem occurs mainly because the available bandwidth varies during the length of the call as people join or drop the conference. The processing power problem is caused by new and more powerful endpoints that send input information which requires greater processing power from the receiver. In both cases, just like with the assembly line example, allowing the receiving side to control the characteristics of the input should help solve the problem efficiently.
Bandwidth estimation in an IP network is prepared by monitoring packet loss indication on the link. Just as the number of dropped pieces of machinery on the floor of the assembly line would serve as indication to a problem with the pace, packets lost (discarded) serve as indication that the available bandwidth was exceeded.
The standard approach to bandwidth adaptation is that the receiver will send a request to the sender, asking for a change of bandwidth (known as “flow control message”). Intelligent bandwidth adaptation algorithms, such as RADVISION’s QualiVision, apply bandwidth estimation and adaptation to optimize the available bandwidth utilization.
The simple solution to the assembly line worker problem has a flaw when it is implemented in the video conferencing world. As the bit rate is being reduced, the quality of the video stream is also reduced. Therefore, the system handles the bandwidth adaptation problem but harms the overall quality.
A better approach is to reduce the resolution (picture size) along with the bit rate, while maintaining the high quality. LifeSize endpoints, for example, change the resolution automatically as a result of flow control messages. This is very beneficial from an endpoint point of view, but causes more instability on a system level – the receiver no longer is in control over the process.
In the past few years, the problem of giving the receiver – and not the sender – the control over the parameters of the video sent to it has been the focus of many in the video conferencing business. An ITU study group (SG16), led by RADVISION, has been designing such a mechanism over the last two years.
The proposal is to add a “Video Sub-Mode Control” (see original draft here – pdf) to H.241, the ITU recommendation that deals with “video procedures and control signals”, in order that better control can be achieved during a call by the receiver of a video stream over parameters such as frame rate (Macro Block rate, to be more accurate), resolution (picture size), frame aspect ratio, etc. All of these can help in both optimizing the use of a given bandwidth and the use of processing power.
Due to the urgency to find a solution to this problem, all leading vendors generally agreed to the proposal, which was consented in this month’s Geneva meeting. Now all that remains is to deploy this mechanism and make sure that in these “modern times” everyone will be able to enjoy high quality unified communications, as we expect and deserve.