Network architectures are based on many factors, including types of applications and protocols, distances, usage and access patterns, and legacy network topologies. In the metropolitan market, for example, point-to-point topologies might be used for connecting enterprise locations, ring topologies for connecting inter-office facilities (IOFs) and for residential access, and mesh topologies might be used for inter-POP connections and connections to the long-haul backbone. In effect, the optical layer must be capable of supporting many topologies and, because of unpredictable developments in this area, those topologies must be flexible.
Today, the main topologies in deployment are point-to-point and ring. With point-to-point links over DWDM between large enterprise sites, there needs only to be a customer premise device for converting application traffic to specific wavelengths and multiplexing. Carriers with linear-ring topologies can evolve toward full rings based on OADMs. As configurable optical cross connects and switches become more common, these point-to-point and ring networks will be interconnected into meshes, transforming optical metropolitan networks into fully flexible platforms.
Point-to-Point Topologies
Point-to-point topologies can be implemented with or without OADM. These networks are characterized by ultra-high channel speeds (10 to 40 Gbps), high signal integrity and reliability, and fast path restoration. In long-haul networks, the distance between transmitter and receiver can be several hundred kilometers, and the number of amplifiers required between endpoints is typically less than 10. In the MAN, amplifiers are often not needed.
Protection in point-to-point topologies can be provided in a couple of ways. In first generation equipment, redundancy is at the system level. Parallel links connect redundant systems at either end. Switchover in case of failure is the responsibility of the client equipment (a switch or router, for example), while the DWDM systems themselves just provide capacity.
In second generation equipment, redundancy is at the card level. Parallel links connect single systems at either end that contain redundant transponders, multiplexers, and CPUs. Here protection has migrated to the DWDM equipment, with switching decisions under local control. One type of implementation, for example, uses a 1 + 1 protection scheme based on SONET Automatic Protection Switching (APS). See Figure 3-8.
Ring Topologies
Rings are the most common architecture found in metropolitan areas and span a few tens of kilometers. The fiber ring might contain as few as four wavelength channels, and typically fewer nodes than channels. Bit rate is in the range of 622 Mbps to 10 Gbps per channel.
Ring configurations can be deployed with one or more DWDM systems, supporting any-to-any traffic, or they can have a hub station and one or more OADM nodes, or satellites (see Figure 3-9). At the hub node traffic originates, is terminated and managed, and connectivity with other networks is established.
At the OADM nodes, selected wavelengths are dropped and added, while the others pass through transparently (express channels). In this way, ring architectures allow nodes on the ring to provide access to network elements such as routers, switches, or servers by adding or dropping wavelength channels in the optical domain. With increase in number of OADMs, however, the signal is subject to loss and amplification can be required.
Candidate networks for DWDM application in the metropolitan area are often already based on SONET ring structures with 1 + 1 fiber protection. Thus schemes such as Unidirectional Path Switched Ring (UPSR) or Bidirectional Line Switched Ring (BLSR) can be reused for DWDM implementations. Figure 3-10 shows a UPSR scheme with two fibers. Here, hub and nodes send on two counter-rotating rings, but the same fiber is normally being used by all equipment to receive the signal; hence the name unidirectional. If the working ring should fail, the receiving equipment switches to the other pair. Although this provides full redundancy to the path, no bandwidth reuse is possible, as the redundant fiber must always be ready to carry the working traffic. This scheme is most commonly used in access networks.
Other schemes, such as Bidirectional Line Switched Ring (BLSR), allow traffic to travel from the sending to the receiving node by the most direct route. Because of this, BLSR is considered preferable for core SONET networks, especially when implemented with four fibers, which offers complete redundancy.
Mesh Topologies
Mesh architectures are the future of optical networks. As networks evolve, rings and point-to-point architectures will still have a place, but mesh promises to be the most robust topology. This development will be enabled by the introduction of configurable optical cross-connects and switches that will in some cases replace and in other cases supplement fixed DWDM devices.
From a design standpoint, there is a graceful evolutionary path available from point-to-point to mesh topologies. By beginning with point-to-point links, equipped with OADM nodes at the outset for flexibility, and subsequently interconnecting them, the network can evolve into a mesh without a complete redesign. Additionally, mesh and ring topologies can be joined by point-to-point links (see Figure 3-11).
DWDM mesh networks, consisting of interconnected all-optical nodes, will require the next generation of protection. Where previous protection schemes relied upon redundancy at the system, card, or fiber level, redundancy will now migrate to the wavelength level. This means, among other things, that a data channel might change wavelengths as it makes its way through the network, due either to routing or to a switch in wavelength because of a fault. The situation is analogous to that of a virtual circuit through an ATM cloud, which can experience changes in its virtual path identifier (VPI)/virtual channel identifier (VCI) values at switching points. In optical networks, this concept is sometimes called a light path.
Mesh networks will therefore require a high degree of intelligence to perform the functions of protection and bandwidth management, including fiber and wavelength switching. The benefits in flexibility and efficiency, however, are potentially great. Fiber usage, which can be low in ring solutions because of the requirement for protection fibers on each ring, can be improved in a mesh design. Protection and restoration can be based on shared paths, thereby requiring fewer fiber pairs for the same amount of traffic and not wasting unused wavelengths.
Finally, mesh networks will be highly dependent upon software for management. A protocol based on Multiprotocol Label Switching (MPLS) is under development to support routed paths through an all-optical network. In addition, network management will require an as-yet unstandardized channel to carry messages among the network elements.
Hernandez Caballero Indiana M. CI: 15.242.745
Asignatura: SCO
Fuente:http://www.cisco.com/application/pdf/en/us/guest/products/ps2011/c2001/ccmigration_09186a00802342cf.pdf
