technology & profits
Choosing the best path to optical-network profits
Should carriers remain lords of the rings or prepare to evolve their core networks to mesh topologies?By Kathleen Richards
If you're suffering from information overload, you can relate to the current state of many communications networks because most carriers today face the same dilemma. It's hard to know just what to do with all that data.
As incumbent long-haul carriers grapple with the snowballing growth of Internet-protocol (IP) traffic, their ring architectures designed to carry voice services are proving inefficient for data transport. IP traffic, forecast to increase to 20 to 30 times its present size in 2002, will change the design and the architecture of the transport network over the next five years, according to Wit SoundView Corp.
The catch-22 is that voice services still generate the majority of these carriers' revenues. So what's the best solution? Build more rings? Wait for the technology pieces to fall in place and migrate to mesh?
Familiar ring
Today, most carriers operate ring-based backbone networks to transport all traffic types and use multiple protocols such as IP, Asynchronous Transfer Mode, Synchronous Optical Network (SONET), and dense wavelength-division multiplexing on different layers to engineer and groom traffic. Although ring architectures were designed to handle voice, incumbent carriers are comfortable with the technology, have huge investments in network equipment, and when additional equipment is needed, it is readily available.In its simplest form, the ring topology is a closed loop with one path for working traffic and another "on standby" for protection, to provide 50-msec restoration if a fault occurs in the working path.
The technology used on rings is also very familiar to carriers. Rings are usually populated with SONET-based systems such as add/drop multiplexers, most often supplied by a single vendor. The operating software built into the systems-usually proprietary-controls the protection switching and the restoration. In theory, multiple vendors' products can sometimes communicate and populate the same ring; in practice, this scenario is rare. Yet, rings have more than adequately served their purpose.
 |
| In a constrained mesh architecture, the network will reroute traffic after a fault or equipment failure on the working path to the best alternate path within the domain. |
|---|
"Rings are ideal for voice-based traffic," observes Dr. Krishna Bala, a member of the Bellcore team that developed ring architecture and now chief technology officer at optical-switch vendor Tellium. Bala says that the locality of voice traffic is the fundamental basis on which ring architectures were built.
"You are more likely to talk to your neighbor if you live in New York, for example, and less likely to talk to somebody in California," he explains. "The traffic patterns are hierarchical in nature. Rings are extremely good architectures for that premise because they are based on the fact that you are multiplexing up into the hierarchy."
This all falls apart when it comes to data, however. And on many carriers networks the data traffic is overtaking voice.
"Data traffic is not hierarchical in nature," says Bala. "It's random, arbitrary, unpredictable, and meshed. Nobody has any idea anymore where the traffic is coming from. When I get on my Web page, and the Web server is in China, I think the data is coming from China. When, it is actually coming from a server in Dallas."
More efficient use of bandwidth
A mesh topology, where nodes are interconnected through multiple paths, is better suited to support data traffic patterns. Packet-based data does not require the constant connection demanded by voice services. A mesh topology in conjunction with IP will allow carriers to make more efficient use of network bandwidth by prioritizing traffic. "A mesh can provide various grades of service, which a ring cannot provide," says Adel Saleh, chief network architect at Corvis Corp.The amount of capacity and fiber required for very-high-reliability protection and restoration schemes in a mesh topology is significantly less than the 100% redundancy required for rings. One reason is that a mesh architecture can provide a connectivity that is richer than a ring-based architecture for a given fiber capacity. Lucent estimates that a mesh architecture for the core backbone would require only 15% to 20% of the fiber for restoration versus the 50% required today for rings.
As traffic explodes, many ring architectures experience fiber exhaust and carriers are then required to stack rings. "The worst nightmare with rings is that if one leg gets exhausted, the whole ring is exhausted," says Bala.
An advantage of mesh is that as carriers start to add capacity, they can upgrade one link, for example, the route with the heaviest traffic, or they can upgrade several links without having to upgrade the entire mesh architecture.
Switch to mesh
The momentum behind mesh as the topology for the next-generation core backbone is driven by more than an overabundance of IP-based data traffic. In the next few years, large optical crossconnects, a key hardware component to enabling protection, restoration, and provisioning in the optical layer using a mesh topology, will become commercially available. Large crossconnects will allow backbone carriers to manage higher numbers of optical signals at their points of presence or nodes. These crossconnects are optical-to-electrical-to-optical (OEO) today, with the largest at 512-by-512 ports. Developers of OEO crossconnects include Tellium, Ciena, Cisco Systems (Monterey Networks), and Sycamore Networks.In two to three years, these large crossconnects will become all-optical, but vendors are finding the technology challenging, according to analysts at Wit Sound View. Developers of all-optical crossconnects include Lucent Technologies, Corvis, Astral Point, and Nortel Networks, through its recent acquisition of Xros. Vendors and carriers are also working to develop software standards to control the signaling and allow interoperability between equipment such as switches and routers. Critical networking issues remain unsolved, however, such as all-optical performance monitoring and wavelength conversion.
"OEO crossconnects with OC-48 or OC-192 cap abilities are starting to emerge," says Corvis's Saleh. "The products avail able are not quite large enough, but vendors are promising that they are going to be larger, possibly 1,000 by 1,000 ports as early as next year. You need a size like this to enable mesh restoration. If you have a small crossconnect, it is hard to gain the full advantage of mesh restoration."
All-optical crossconnects, which eliminate the need for the optical-to-electrical and electrical-to-optical conversions are on the horizon. "With the all optical crossconnects, the capacity of the switch is much larger," notes Saleh.
While many associate the emergence of the large crossconnects as the key enabler of a physical mesh topology, vendors such as Nortel view the topology debate as primarily a restoration issue.
"Optical crossconnects imply connection management capability," asserts James Frodsham, vice president of product-line management at Nortel's optical-networks group. "Mesh versus ring are types of restoration, and optical crossconnects support both types of restoration.
"What carriers want is flexible connection management capability with the ability to provision class of service on a connection basis," adds Frodsham. "Part of service provisioning flexibility will entail providing them with the ability to select mesh-based restoration, or ring-based restoration, or a non-protected, non-restorable connection."
While in theory, a mesh topology offers alternative paths for switching traffic when a fault or equipment failure occurs on the working path, in practice, that's not the case in a lot of networks. In addition, even if another fiber route is available from that junction, the actual route that traffic traverses may end up being significantly longer than the alternate path the ring offered; that all depends on the richness of the fiber connectivity that the carrier has. The mesh topology may offer greater efficiency, but the cost of that efficiency if it causes carriers to restore traffic down an extremely long restoration path might actually drive up the cost in the network.
"In a perfect world, mesh wins out every time," agrees Frodsham, "but in a practical world, the economics of a bounded mesh construct will probably offer the optimum balance between the existing ring technologies and the sort of hypothetical optimum mesh topology." How protection and restoration work in a ring, mesh, and constrained or "bounded" mesh architecture is shown in the Figure.
Getting from here to there
Emerging data carriers such as Extant Communications (Aurora, CO), which uses Tellium's optical switches and software management system, and Qwest Communications, which uses Corvis's products, are beginning to build mesh architectures. But how should incumbent carriers with extensive ring topologies incorporate mesh architectures?Today, most carriers should be investigating how mesh topologies will work as well as learning about the emerging equipment and software that will support these architectures. "If I were a carrier, I would keep my rings as they are and start deploying mesh in small segments of the network as soon as I could get the products and the software to become familiar with the technology," says Tellium's Bala. "An advantage of mesh is that it can be built arbitrarily, you don't have to plan the entire network over the next five years, you can drop these products as you will in the network."
In Bala's view, the voice traffic will remain on the rings and the data traffic will eventually migrate completely to a meshed architecture. "If you start going to mesh today, within about three years, you will have about five times the infrastructure that you have today. So today's network won't be significant, it will be too small to worry about."
Another architecture that is emerging is a two-tier architecture, according to Corvis's Saleh. "The amount of traffic growth expected in the backbone is enormous, so people are not likely to dismantle the rings and remove the SONET add/drop multiplexers and put in the mesh instead," says Saleh. "On each route incumbent carriers may have from a handful to a hundred fibers in the ground, so they are likely to build an overlay mesh network and use the existing rings as regional collecting rings passed by multiple speeds.
"The ring did not start as the national backbone network method," observes Saleh. "It started in the metropolitan area and they extended it to the regional network and then to the backbone. It is now going back to the regional network."
Rings are not going away. Some carriers will continue to build rings, based in part on the current availability of the network equipment. And there may be even more rings in the future, as ring topology is used when providers build-out regional and access networks. "It may not be SONET add/drop multiplexer-based rings as we know them," says Saleh. Future rings will be equipped with small crossconnects designed to emulate SONET add/drop multiplexers. A few vendors are already developing these products.