March 13, 2008

IPTV Thought Experiments

By Jonathan Tombes

The "IP Cable" track (one of six) at the IPTV World Forum taking place in London this week was the occasion for some forward thinking on some potentially powerful benefits of Internet protocol (IP) transmission of video.

In a keynote address on Wednesday's opening session, Cisco Systems' S.V. Vasudevan took a slightly quirky seven (rather than five or 10) year gander at the efficiencies of a variable bit rate (VBR)-friendly, channel-bonded future.

Nothing so peculiar about the actual statistics of this long-necked view, which was corroborated by separate research shared by a competitor, ARRIS Group VP Advanced Technology Mark Bugajski, in a Thursday morning session.

Getting to this future point does entail jumping over a daunting reality that Vasudevan gamely noted up front, namely: that existing IPTV solutions for cable both increase cost and diminish throughput.

Never mind. Let's assume the trends are IP's friends. Vasudevan said that those include steady growth of telco-based IPTV customers (e.g. French provider Free already has more than 2 million subs); reduction in size of video assets, thanks to MPEG-4 AVC; and widespread proliferation of both on-demand and DOCSIS 3.0 technologies.

From that point, Vasudevan spun off two well-known (if not incompatible) alternatives: the proliferation switched digital video and channel bonding. The former turns critical points in the cable network into the equivalent of the Chicago stockyards (characterized by acres of precisely spaced tracks, or channels) with the edge resource manager (ERM) playing the role of the switchman. The latter turns cable transport into the opto-electronic version of the Alaska pipeline.

Let's assume

Now, let's further assume not simply the three, four or eight channel bonded options that currently exist, but something more. Vasudevan suggested 24 by 2015. That's far-fetched, but not unreasonable, as existing spectrum would allow for about 125.

Here's the kicker: The Gigabit to the home enabled by 24 bonded channels (40 Mbps of 1 QAM channel x 24 = 0.96 Gig) would allow all video services to travel in an untouched VBR format, and ensuing statistical efficiencies add up to a 30 to 40 percent win over an alternative approach using constant bit rate (CBR) encoding.

For his part, Bugajski brought that insight back into existing two, three and four channel bonded approaches. The results of a study of supported by the European Commission and conducted by Polish academics indicate more than a 40 percent bandwidth efficiency of VBR and CBR methods, without the drawbacks to video quality of CBR.

The statistical benefits grow as throughput increases. At 100 Mbps, the gain is five extra movies; at 150 Mbps, the gain is eight; and at 200 Mbps, 12 movies. "As you add more, the probability of the I-frames colliding gets smaller and smaller," Bugajski said.

Vendors with a stake in CBR and an eye on the near-term competitive realities, especially in North America, where countering the direct broadcast satellite (DBS) high-definition television (HDTV) assault has turned into trench warfare, may offer another view of these numbers. These are interesting numbers, all the same.

- Jonathan Tombes