Network survivability top concern for architects
Network survivability top concern for architects
Probe Research Inc.
During the next few years, network survivability will be a key concern of network architects (in the cable-TV, local exchange and electrical utility industries) who work with hybrid fiber/coaxial-cable technology, and expect to deliver a full suite of services, ranging from entertainment video and multimedia to telephony. In particular, architects will be looking for ways to achieve the standard network availability of 99.99 percent (no more than 53 minutes of total network outage per customer per year) now expected from local telephone networks. As part of that effort, planners increasingly will make extensive use of diverse fiber routing and redundancy in key opto-electronic components
Still, the use of coaxial cable for the network "tails" will continue to make financial sense, requiring judgments about the amount of investment a carrier is willing to make to improve network availability for different classes of service. Although the elegant answer is deployment of underground fiber to every residential premises, that is almost certainly the wrong financial answer for service providers anticipating fierce and widespread local loop competition in the latter 1990s.
To create a baseline, consider a variety of early hybrid fiber/coaxial- cable networks that progressively deploy more optical fiber in the local loop. In January 1992, technologists from Cox Cable Communications Inc., one of the largest U.S. cable-TV companies, reported on reliability improvements seen in Cox`s U.S. networks once optical fiber was introduced. Those figures indicate that a modern hybrid fiber/ coaxial-cable network will approach the standard recommended for local access networks.
For a network using fiber to an optical network unit or "node," and featuring the use of four additional radio frequency amplifiers in series, network availability is in the 99.995 percent range. At that level, any customer might expect to experience approximately 25 minutes per year of network downtime, a level good enough to meet consumer expectations for telephony service.
In general, methods to improve hybrid fiber/coaxial-cable network availability include the following steps:
Redundant opto-electronics, transmit and receive
Minimize active points of failure
Use telemetry to monitor network
Maintain accurate plant documentation
Reduce size of optical serving area
Power supply location on utility grid (upstream of fuses, surge protection)
Uninterruptible power supplies
Reduce size of plant that is fed from any one power supply
Use surge protection.
Although these steps help, route diversity is quite important. The rule of thumb for U.S. cable-TV operators is that fiber breaks are the largest single cause of network downtime. And, according to technologists at Rogers Cable Communications, the largest Canadian cable-TV operator, a reasonable expectation of optical fiber plant failure is that 58 percent of total failures are caused by fiber cuts, in particular "dig-ups," which sever underground cables.
Reliability of 99.67 percent might be good enough for standard cable-TV service, suggest officials at Cable Television Laboratories, the research and development arm of the U.S. cable-TV industry. However, Rogers officials suggest that information-on-demand and video-on-demand services might require a 99.98 percent availability standard (105 minutes outage per year).
Looking at various hybrid fiber/coaxial-cable-based local access topologies, Rogers engineers used internal benchmarks for fiber and amplifier failure (fiber, 1.5 percent; amplifiers, 5 percent per year) to measure performance against the hypothetical target of 99.9999 percent and a real-world figure of 99.985 for star-star design local telephone plant. That analysis suggests that a hybrid fiber/coaxial-cable network using three radio frequency amplifiers in series, after the ONU, will achieve availability of approximately 99.98 percent, translating to 105 minutes per year of downtime. Nevertheless, Rogers engineers point out, that is a standard deemed adequate for DS-1 service (1.544 megabits per second), and better than the 99.97 percent availability specified for inter-LATA DS-1 service.
To this point, the dramatic gain in network availability for a network with 24 amplifiers (it has in the past been common for U.S. and Canadian cable-TV networks to run 30 to 35 amplifiers in series) has been in the move from all-copper design to hybrid fiber/coaxial cable with four radio-frequency amplifiers, reducing outages from approximately 2.5 hours to 25 minutes or so (on average, about a 6:1 improvement), according to Cox Cable figures.
In fact, to this point, neither local exchange carrier nor cable-TV executives adopting hybrid fiber/coaxial cable as a broadband access platform have believed the additional cost of route diversity between headends/ central offices and ONUs/nodes was warranted. However, with the advent of competitive local telecommunications, hybrid fiber/coaxial-cable architects have begun to think more seriously about methods for further increasing the survivability of hybrid fiber/coaxial-cable networks.
Looking just at the outside plant and opto-electronics that terminate the outside plant, redundancy options traditionally fall into a few categories:
Component redundancy involves backup of the primary component or network element with a spare unit, while fiber path backup normally would take the form of diverse routing, so no single cable cut can sever multiple fibers in the sheath. This approach is highly survivable, but also twice as expensive as the non-protected line.
An alternate deployment that offers protection from fiber cuts but lowers cost would retain route diversity but eliminate the use of redundant transmitters, while continuing to back up the transmitters. This protects the network from cable cuts or receiver failures, at a lower cost than the full-redundancy deployment; but it is susceptible to transmitter failures. Note that a headend transmitter failure is less a problem than a transmitter failure in the outside plant, because rack-mount units typically can be replaced relatively quickly. A less-robust option would keep diverse fiber-path routing, without backup of the opto-electronics. In this case, optical splitting or switching is used to select from the primary and backup paths.
When using an optical splitter, the light from the transmitter is fed to two outputs, which are sent over diverse routes. At the receiver location, a radio-frequency switch selects which of the two inputs is the primary and which is a protection feed. One problem with this approach, says Richard Roth, OptiVideo president, is that "the transmitter power must be increased by a factor of two or more" to feed both the primary and protection fibers. "That`s essentially wasted light," he says.
One way of dealing with that problem is to substitute an optical switch for a splitter. When using an optical switch, transmitter light is sent only along the primary path, unless a fault condition occurs. Then the light is sent along the protection route. That has one significant advantage, says Roth. It eliminates the need for generating twice as much light.
That leads to a new problem: "A fiber not carrying light from end-to-end may be damaged and might go undetected," says Roth, whose firm supplies optical switches to cable-TV and telephone companies. To overcome that particular difficulty, he suggests service providers use a switchable splitter, an active optical coupler featuring a variable splitting ratio, "set to obtain a small number of distinct states, usually four or less, with split ratios ranging from 0/100 to 100/0."
The idea is to send most of the light down the primary path, injecting just enough light into the protection fiber so it can be detected and measured on the other end of the path. That way, the carrier always knows the protection path is intact and functional, while at the same time avoiding the need for additional optical power, says Roth.
By placing a 2ٴ optical switch at each receiver, and a switchable splitter in front of each transmitter, a carrier might retain all the advantages of route diversity while reducing redundant transceiver costs, while simultaneously keeping ability to test the link for integrity, says Roth.
Although full component protection and route diversity provides the most protection against cable cuts and device failures, the cost penalty can be quite serious. Consider the case of a single link, of 18.5 kilometers in length (approximately 11.5 miles), running four fibers in a single cable. As a conservative estimate, full path redundancy for this link would run at least $288,075 or about $25,050 a mile. Even though construction, optical fiber and cabling costs for any diverse routing of this type will be constant for any opto-electronics protection scheme, the potential savings come in the area of a transmitter and receiver pair, at approximately $32,000.
This analysis assumes a star-type route diversity plan in which both sets of diverse cables run directly between a headend/central office and an ONU/node. Other configurations are conceivable, including ring designs that link nodes back to the headend in a sort of clamshell design, or a simpler "pie wedge" design in which each set of two adjacent nodes are connected.
The cost differential for the pie wedge design may be slightly less than for the star-type design. In this case, where the cross-connected nodes are 3.25 miles apart, the additional cost of fiber, transmitters and receivers is approximately $105,538.
As the year progresses, we are likely to see continued refinement of route diversity and opto-electronics protection schemes, driven in part by the desire to reduce cost, and partly by new thinking about how protection best can be provided. At the same time, thinking about transport platforms for a variety of new services--focusing on bandwidth flexibility--will combine with the new wave of thinking about topologies. The result surely will be a tsunami of new thinking about hybrid fiber coaxial-cable platforms.
Gary Kim is senior vice president at Probe Research Inc. in Cedar Knolls, NJ.