Achieving 99.99 percent availability
Some excellent papers on network reliability were presented at the NCTA Technical sessions in Dallas last May, and on network availability at the SCTE Conference on Emerging Technologies in San Francisco in January. Engineers and technicians responsible for planning and operating broadband networks need to become familiar with the concepts, as well as the causes and remedies, discussed in these solidly professional papers. The opportunities unleashed by the signing of the Telecommunications Act of 1996 impress upon broadband networks generally an enormous urgency to achieve dependable upstream transmission.
The authors of the papers make an important distinction between "reliability," a commonly misunderstood term, and "availability," the term which better represents what we really mean. Reliability is the probability that the network will perform acceptably, without repairs, over a specified period of time. Availability, on the other hand, is the percent-ratio of the time during which the network performs acceptably, either on average, or for individual cases, to the total time, including downtime. On the average, this is the ratio of "mean time between failures (MTBF)" to the sum of MTBF and the "mean time to repair, or restore, service (MTTR)."
In the 1989 Viacom Customer Service Survey (ONTRAQ), "excellent" customer ratings were reported for service with fewer than two total outages in a three-month period, equivalent to eight outages in a 12-month period. This survey was incorporated in the report of the CableLabs Outage Reduction Task Force in 1992. For an excellent rating, then, MTBF should be at least one-eighth of 8,760, or 1,095 hours. If four hours are required to restore service, including main power failure, as well as faults at the headend, distribution network, and CPE, as suggested by CableLabs, Rogers Cable, and others, the availability would be 1,095/(1,095+4) = 0.9964 (i.e., 99.64 percent).
Certainly, competition will demand better than eight outages per year, lasting four hours each (0.9964 availability). For an HFC network with properly monitored standby power supplies, a good status monitoring system and redundant headend facilities, assuring that no subscriber would experience more than one or two total outages per year would not seem an unreasonable objective. Even so, assuming four-hour MTTR, the availability would be only 0.9991 to 0.9995. The arithmetic illustrated in the accompanying chart leads to the conclusion that achieving the last few hundredths of a percentage point will be difficult and expensive. It will require either a substantial increase in MTBF or a substantial decrease in MTTR, or both.Telephony vs. video requirements
Competition from DBS, digital MMDS, and perhaps telephone video dialtone (VDT) or its successor is likely to compel better performance than represented by two outages in three months' time. Technology for the standby power and performance status monitoring needed to reduce the frequency and duration of outages is at hand. However, there is no reason to believe that the video entertainment competitors are likely to achieve the "four 9s," nor does there seem to be credible economic motivation to do so.
On the other hand, the "four 9s" is not an unreasonable objective for telephone service, whether it be lifeline POTS or more sophisticated high-speed data service. Availability is critical, not only in life-threatening situations, but also in all but the most casual commercial and personal communications.
It ought to be feasible to consider the availability requirement for telephony independently from the requirement for entertainment video distribution. Separate upstream coaxial cables would provide more bandwidth and more flexibility in frequency management than is possible in the 5–40 MHz FDM architecture. "Sruki" Switzer told the 1995 NCTA Convention in Dallas that: "The present 5–40 MHz sub-low reverse path . . . is like building a super-highway with eight lanes in one direction, and a dirt track in the other." Using 900 to 1000 MHz for reverse transmission could obstruct future upgrade and development much as HRC has done. Moreover, separate coaxial cables for telephone and video between user premises and the optical node could help to concentrate the money and effort where it is most needed to achieve the four 9s.
The incremental cost of separate coaxial cables might be at least partially offset by the lower cost of a one-way video network. Remember the "AquaCar"? It was neither a good automobile nor a good boat. Would an integrated coaxial network turn out to be less than the best for either video entertainment or modern telephony? Think about it.