DOCSIS, the Data Over Cable Systems Interface Specification, was brought about for one very logical and simple reason: to provide a standard to allow for interoperability between the equipment of different vendors. Through DOCSIS, MSOs would be able to deploy services faster due to there being DOCSIS-compatible cable modems available at a plethora of retail locations.
Customers could just install their compliant cable modem themselves without their MSO having to send out one or more technicians to do it. In addition, the standards would allow for the quicker adoption of multiple services (i.e. voice, data, video-on-demand and other digital television options) through both cable modems and set top boxes. Indeed, many manufacturers are merging those two boxes into one, full-featured device. That convergence that everyone in the industry has been banking on and praying for over the years, would finally become a reality thanks, in part, to DOCSIS. The specification isn't without its problems, however.
What I see as the biggest problem with DOCSIS is its severe lack of upstream channel bandwidth. Whereas most of the pre-DOCSIS devices made use of two 6 MHz channels- one for upstream transmission and one for downstream, the standard specifies a maximum upstream channel bandwidth of only 3.2 MHz. In fact, a reverse path as low as 200 KHz can be used.
The downstream channel is the standard size of 6 MHz. To compliment this part of the spec, DOCSIS requires that the upstream channel lie in the noisy 5 - 42 MHz spectrum and that either Q-PSK (quadrature phase shift keying) or 16-QAM (quadrature amplitude modulation) modulation schemes can be used to facilitate it. Once again, the downstream channel is relatively unaffected by the spec and can lie anywhere between 88 and 860 MHz.
The fact that the size of a data segment is determined by the amount of subscribers who make use of the upstream -- not downstream -- channel at any given point in time, really hurts the DOCSIS design. It is the upstream channel that is shared between those subscribers because they must contend for its capacity any time they use their connections. The downstream channel is dedicated to the gateway router at the regional headend, so that is not a bottleneck at all. True, the current model of Internet usage is for highly asymmetric applications like web browsing, but even then upstream capacity is used any time a connection is open.
What, then, would the effect be of simply cutting upstream channel bandwidth by at least half, as specified by DOCSIS? Logic dictates that a maximum of half of the original subscribers populated per data segment would be able to be supported using the DOCSIS architecture. That means that MSOs would have to go on a segment-splitting frenzy in order to properly implement the compliant hardware. Otherwise subscribers would immediately realize excessive packet loss and latency, which would contribute to horrible overall network performance.
There would also have to be very stringent upstream usage guidelines specified for subscribers, in order to prevent saturation of the reverse path. That means implementing a total ban on servers, as casual as they might be, which I would find to be highly inappropriate and draconian with respect to subscriber usage. There is no free lunch inherent in chopping upstream channel capacity in half, or more, and the sooner MSOs realize this, the better.
Something else that can severely hinder DOCSIS in the future is the never-ending impact of ingress noise on the reverse path. This is and has always been cable's most limiting factor with respect to deployment of data services. A fatal flaw, if you will, just like DSL has its distance limitations. The biggest problem here is that in order to obtain an upstream data rate of 10 Megabits per second with DOCSIS, which is what's being used by many operators at the moment, 16-QAM must be implemented as the modulation scheme.
Using that kind of modulation over a part of spectrum that has traditionally been very noisy, is no less than a recipe for disaster. Use the more robust Q-PSK and you only get half the bit rate, and thus half the subscribers per data segment. I might emphasize that no matter how low the noise floor is kept and how clean an MSO's plant is, noise will always creep in at the subscriber drops. The only way to combat this is to use more robust modulation schemes (Q-PSK is just fine for this task) with 6 MHz channels, in a more hospitable spectrum.
For instance, the last "A" channel, 108-114 MHz, is currently being used by Rogers to provision my upstream channel. Why can't others use it as well? Techs should also be sent out at a moment's notice to check subscribers' drops if there are no outstanding hub or trunking problems in their areas. In all my experiences with consumers, I have only seen this done a couple of times, which I find to be abominable. If I encounter problems with my television reception, getting a tech visit is quick and simple. Why the double standard for data services; and why is such a burden of proof placed on subscribers to prove that they have any problems in the first place?
The DOCSIS standards were first created with some very legitimate, logical and honorable intentions in mind. When the first specs were being drawn up, actual cable modem subscribers were few and far between. Operators and vendors alike worried incessantly about the prospects for future speedy deployment and the encroachment of DSL. DOCSIS made a lot of sense back then because of those reasons.
However, the passage of time has shown us that some of the oldest solutions are the most effective in the real world (for example, bridged Ethernet over a 10BROAD36 architecture, as in how the LANCity operates) and that the RBOCs are trailing badly behind cable in provisioning competing DSL services.
If current DOCSIS specifications aren't amended, though, that could very well allow telcos to catch up to cable with DSL in the long term. The surefire way for cable to win is to outmatch DSL in throughput, latency and reliability, which can be done in time.
Cable, at its most basic level, has an inherent latency and throughput advantage over DSL, which should be fully leveraged to gain further market dominance and keep it ahead of DSL in the broadband race. It is clear that current DOCSIS specs aren't built for performance or scalability, particularly if usage becomes more symmetric, and that can cost both cable operators and consumers quite a bit in the long term.
Christopher Weisdorf is the president of the Rogers@Home Users Association.