Implementing redundant fiber architectures
As system operators offer advanced services such as video-on-demand, Internet access and telephony, the need for providing reliable, "interrupt-free" service escalates. The first step to ensuring that customers receive high-quality, reliable service is to implement redundant network architectures, where a backup system seamlessly fills in during times of primary system outage. But redundancy alone is not a panacea; although redundant networks provide enhanced network reliability, they bring their own pitfalls. By spending the necessary effort before implementing redundant configurations, operators can save resources while still providing reliable, high-quality service.
Operators must first select what level of system redundancy to implement. In their fullest extent, backup systems replicate primary systems entirely, with a fully automatic switch detecting any loss of service in the primary route and switching the transmission path to the backup architecture within milliseconds. Although this configuration provides the highest coverage, the initial costs of the primary infrastructure are virtually doubled because every element of the primary path has an identical backup component.
Because many components in a given fiber link are extremely reliable (e.g., the mean time between failure of a distributed feedback laser used in a typical transmitter is over 30 years), providing an entire backup system is not always necessary. Many services do not require switching to backup systems within milliseconds. Unless critical services are being transmitted, operators should also consider methods to restore service within minutes of detected outages. Such redundant systems are much less expensive. Operators can save substantial dollars by designing a redundant system that duplicates only the least reliable components.
The most common causes of service outages are powering problems with the active network elements (such as a utility power outage, a power supply failure, or simply a blown fuse), and secondly, fiber breaks. Together, these account for almost all service outages. By protecting against these failures, system operators can ensure overall network reliability in all but a handful of instances.Power redundancy
To protect against powering outages, redundant powering is necessary. Because both headend and hub locations have high concentrations of equipment in a controlled environment, it is simple and economical to provide a 24 V or 48 V battery and/or generator backup at each site. Optical nodes, alternately, are highly dispersed throughout the network, so installing battery or generator backups in each location is costly. An economical way to provide redundant powering of nodes is to install dual power supplies within the node, each supplied by an AC supply bus connected to different sections of the power grid. Each node must have at least two power-ready ports.
Dual powering requires careful implementation for successful redundancy. The dual-power system must:
- Be self-starting. When AC power is applied to either the primary or backup bus, the node must begin operating.
- Automatically switch to backup power if the AC power on the primary bus fails.
- Automatically switch back to primary power when AC power on the primary bus is restored.
- Automatically switch to backup power if the primary power supply fails.
- Automatically switch back to primary power when the primary supply is restored.
Because common switching power supplies are highly efficient at or near full load, but decrease markedly in efficiency near half-load, the dual power supplies must switch in and out, rather than share the power load.
Figure 1 shows a diagram of a circuit whose operations conform to these requirements.
The control circuit shown in Figure 1 contains two small power supplies, one powered by bus 1 and the other from bus 2. The DC output from these supplies is combined with diode protection and used to power the control circuit only. The AC voltage on each bus, as well as the DC output of power supply 1, is also monitored. When either the AC power on bus 1 or the DC output of power supply 1 is not detected, the controller switches to bus 2/power supply 2.
Remote monitoring must monitor critical circuits within the control electronics. Though not likely, a possible sequence of failures could result in an outage which would require field service to restore operation. For example, if the power supply from the backup source supplying the control circuit should fail, concurrent with a primary power failure, the node will not have power, resulting in an outage. In this case, the remote monitoring would notify the operator of a failure within the control unit. The control unit could then be repaired at a convenient time.Route redundancy
To ensure reliable transmissions, data networks such as Sonet use a bi-directional ring with retransmission at each point. This architecture requires only two fibers to connect each point on the ring with its neighbors. At each point, the data is received from the neighboring points and retransmitted both clockwise and counterclockwise to the next point in the ring. When either or both transmissions reach the destination, logic determines when identical data has been received, preventing endless looping of the data.
While this method is effective for digital transmission—where repeats are lossless—it is impractical for analog transmission, with its high performance penalty for repeating. To achieve high performance, analog transmission networks such as cable TV systems must have both direct primary and backup routes from the source to each destination through a minimum of repeats. As shown in Figure 2, a redundant fiber route allows for two parallel transport systems. When the primary system fails, the backup system can be selected. In this manner, operators can restore service more quickly than repairing the primary system.
As in digital systems, the most efficient method of implementing redundant routes to multiple locations is by using rings. A ring with three destinations and one source, or four points, is shown in Figure 3. Two routes primary and backup from the source to each destination are required.
In general, a redundant ring for the forward path of a cable TV system with one source and (n-1) destinations, or (n) points, must be completely encircled with (n-1) fibers. The return path functions in much the same manner, with (n-1) sources and one destination. The return path also requires (n-1) fibers completely encircling the ring.
Because of the distances involved, it can be difficult to achieve high performance on the backup links of large rings. One solution is to accept slightly poorer performance on the backup routes than on the primary routes. Another possibility is to use one or two repeat sites at or near the far end of large rings.
Once the operator has determined the redundant network topology, there are many possible fiber redundant architectures from which to choose. One option is to use an optical switch immediately before the optical receiver. Optical switches, however, have their own drawbacks, such as a high insertion loss which can negatively impact the carrier-to-noise ratio (CNR) of the optical link. In addition to quality degradation, optical switches can be costly.
Another option is to use two optical receivers followed by an RF switch, placing the switch before the main RF amplifier in the receiver location. While placing the switch after the amplifier would result in a redundant amplifier as well as a redundant optical route, the RF amplifier is a reliable component, and very little overall increase in reliability is gained by having a redundant amplifier. As the insertion loss of the RF switch minimally affects the CNR of the link, this method avoids the quality concerns presented by the method above. Also, the cost of the extra receiver and RF switch is generally about the same as that of an optical switch.
A complete, active, route redundant fiber network including both forward and return paths as well as a supertrunk link followed by a distribution link is diagrammed in Figures 4, 5 and 6. Operators can scale this configuration to work with many networks. In this configuration, a pair of transmitters and a pair of receivers on each link, both for the forward and return paths, provide route redundancy. Switching to the backup route is accomplished via two methods, and in either case can take less than 50 milliseconds (ms) if the system is designed properly.
Forward path switching is accomplished via an RF A/B switch. The return path receiver RF outputs are combined, and the route is switched by internally muting the unused receiver. While both of these methods are effective, each has its advantages and disadvantages. Using a combiner and muting a receiver results in 4 dB greater loss in the RF path, while using an A/B switch is slightly more complex and expensive. In general, RF level is less of a concern in the return path. Therefore, an A/B switch is preferable for the forward path, while a combiner is generally preferable for the return.
As supertrunk links typically support hundreds or thousands of times more subscribers than distribution links, operators should make supertrunk links reliable by installing full, active redundancy for all supertrunk links where reliability is a concern. Because distribution links serve many fewer subscribers, it may be practical to install a network to restore service within several minutes (cold backup) rather than 50 ms (active backup). Implementing cold backup can save significant investment, while having minor effects on overall network reliability for links serving few subscribers.
The most cost-effective cold backup method is to install the backup fiber route, backup receiver and switch, without backup transmitter. The backup fiber ends in an optical connector ready to be plugged into a transmitter. If a fiber failure occurs, the network management system notifies the operator, who then plugs the backup fiber into a suitable transmitter. Connecting the backup fiber restores service. This method is effective only if an operator is physically available to make the connection. If the hub location is unmanned, operators must consider the time to travel to the hub site when calculating service restoration time.
With careful planning, operators can scale redundancy to meet their individual network reliability requirements. In many cases today, only supertrunk redundancy is needed. Later, when services demand greater reliability, cold backup on the distribution links can be installed. When even greater reliability is required, active backup can be implemented.Putting it all together
Redundant system topologies can only go so far in increasing system reliability. One of the largest pitfalls in implementation of redundant architectures is the "Jurassic Park syndrome" having the backup system kick in as planned when the primary system fails, but then not realizing that the system is then operating without backup. When the primary system fails and the backup system is in use, the network is no longer redundant. A subsequent failure in the secondary system would result in a service outage.
In order to avoid this undesirable situation, operators must implement a process for monitoring the network, notifying operators that the primary system has failed, and then restoring the primary system before a second failure occurs. A redundant system without these processes cannot provide the needed level of reliability; system failures have not been eradicated, but only postponed until both primary and backup configurations fail.
It is a given that operators need reliable networks to support advanced services, and redundant system architectures are an ideal way of increasing the reliability of service. While there are many options for implementing redundant configurations, operators must consider their individual networks' current and future needs before embarking on creation of a backup infrastructure. With careful planning and modest effort upfront, forward-thinking system operators can develop reliable, high-quality, yet cost-effective networks that will carry them into the next phase of broadband services.