Cablecaster, March 2001

Consumer Window

Hold the phone

Cable telephony: To packet switch, or not to packet switch?

By Christopher Weisdorf, Rogers @Home Users Association

This year marks the 125th anniversary of the invention of the telephone. As such, I could think of no better time to focus on one of the greatest inventions and how more recently, it is affecting cable operators.

Since the early 1990's, much in the way of resources has been devoted to getting telephone service to run over the coaxial cable plant. The company formerly known as Bell Labs -- now Lucent -- made the first attempt, but were forced to cease their efforts due to budgetary concerns. Many others followed suit with their own R&D; and most of met with the same fate as Bell's early efforts.

Fast forward towards the end of the '90's. The Internet has been adopted as a communication medium by millions, bringing with it rapid growth in all areas of technology. Along with that came CableLabs and its DOCSIS solution, where the inevitable questions of whether or not the Internet protocol (IP) could be used to carry voice traffic over cable arose. Hopes, as well as the stock market, were running very high and a great deal of time and money were spent upgrading the coaxial infrastructure. Premiums of thousands of dollars per subscriber were paid in cash and stock buyouts of cable properties, on the expectation that the implementation of Internet, voice and other services would make it all worthwhile in the end.

These days, things aren't looking as bright. Reports of component overcapacity abound, expectations are down across the board, and the sell-off of anything tech-related in the equity markets has been unmerciful. The premium once placed on the tech sector has melted like a slice of butter in a hot frying pan.

Naturally, with MSOs being part and parcel of the technology pie, they face similar difficulties. Their hyped plans for voice over IP have been shelved for the time being, with VidČotron being the first to all but abandon it shortly after its proposed takeover by Quebecor. Most of their North American counterparts have taken the same path, citing 2002 as the earliest year for VoIP deployment. Unfortunately, with things running their present course, that sounds about as believable as the "second half recovery" theory.

Now, with the immediate focus off of VoIP, MSOs can tend to existing issues related to their Internet service and digital television services. It is a good time to reassess the future of telephony over the cable infrastructure, while dealing with the services already out there.

There are two primary problems with VoIP at the moment: Reliability and the natural contention that occurs over the cable network.

Can VoIP be made to be anywhere near as reliable as plain, old telephone service (POTS)? This is a difficult question to answer, but it leans towards the "no" side. Peter Cochrane, British Telecom's chief technologist, agrees. In a 1999 interview, when queried on the topic of migration from circuit switched to packet switched networks, he was blunt.

"There's not a hope in hell that IP [Internet protocol] or ATM [asynchronous transfer mode] will replace and provide the service quality that you get in a phone network. It will never happen," he said.

In the same interview, Cochrane talked about what this space has mentioned before -- that the key parameter in communications is delay. With circuit switching, the delay is constant. With packet switching, however, delay can only be constant if the routing path is fixed and there is always a sufficient amount of network resources available to sustain the connection.

The whole idea of a fixed path flies in the face of the Internet protocol, since it was founded on the premise that the network topology is dynamic and constantly in flux.

With respect to reservation of network resources to provide a set quality of service (QoS), there are a number of protocols which can handle that duty. The real problem is in managing it. When one or more applications need a certain guarantee of bandwidth, reliability and delay, what other applications must function on reduced resources? What services must suffer in order for others to function normally? What is the least amount of resources a particular application can function normally on?

The existence of these questions is one of the main reasons why QoS provisioning doesn't seem right. Just managing a network is difficult enough. QoS muddies the waters substantially and places an undue burden on the person required to answer the above questions. Quality of service is a nice tool to have at your disposal, but its implementation should always be the exception and not the norm.

Addressing the problem of contention (i.e. the sharing that occurs whenever one wishes to transmit onto the cable network), is all reliant upon the modulation scheme(s) in place and how much upstream spectrum is allocated. The latter factor will determine how large data segments can be and how many subscribers can be active at one time. The problem is that a bandwidth-starved scheme such as the one implemented in DOCSIS, doesn't leave much room for anything other than provisioning an internet service. Sharing a 3.1 MHz channel (or less) with both variable bit rate data traffic and constant bit rate voice traffic, is a recipe for disaster. Why? Because during busy periods on the data segment, the QoS features inherent in DOCSIS will ensure that voice traffic flows cleanly -- at the expense of the Internet service sharing the pipe. This would surely not be one of the best ways for an MSO to keep their Internet subscriber base happy.

A good solution to the contention problem would be to introduce a VoIP service to its very own spectrum. One standard 6 MHz channel would likely be more than sufficient to accommodate the needs of even the most aggressive MSOs. The main point here is that the Internet service operates within an upstream channel that is completely separate from that which the voice service operates. Both data and voice traffic can share the downstream channel, but for good measure it would probably be best to completely segregate the two. It's less complicated in the long run.

In pursuit of a cable telephony solution, one cannot ignore the tried-and-true benefits of circuit switching. EastLink, Canada's 7th largest MSO, was the first in North America to implement a circuit switching telephony service over its infrastructure. AT&T; Broadband and Cox also deployed proprietary, circuit switched voice services, albeit to a more limited extent. The real story is EastLink and their telephone service which has now been operating for more than one year. Their claim of 99.99% reliability is very impressive. Even more impressive is the fact that I tried my damnedest to find a negative comment about the service, but was unable to. In lieu of the negative publicity surrounding certain North American MSOs' Internet services and the absence of any real local phone competition, this is encouraging.

In the end, even with a circuit switching implementation, cable operators may never be able to attain the lofty "five nine", or 99.999% reliability of the telephone companies. If 99.99% reliability is the next best thing, so be it. Just like with the traditional circuit switched network, it should be able to operate during power outages. This feature alone is enough to outmatch the reliability of VoIP.

Voice over IP is the new kid on the block. It's still a very immature as a technology and, from its limited trials, needs to undergo much more testing before being deployed on a large scale. Circuit switched telephony over cable can be done today with good reliability and offered at a discounted cost. As a consumer faced with an uncertain macroeconomic outlook, I would welcome this competition now maybe more than ever.

Christopher Weisdorf is president and technical Director of the Rogers @Home Users Association.

Back to Contents

Copyright Notice Copyright © 2002 Business Information Group. All rights reserved.