The optical and terabit route to packet-based networks

Th Acf188


Terabit switch routers can coexist with optical switches, each technology leveraging the other, as carriers migrate to IP networks.

It's no secret that a clock is ticking in the networking world. It's inevitable. In an effort to compete with ISPs, carriers will have to migrate from circuit-switched infrastructures to networks built around packet switching. It's a question of when. When should carriers with substantial investments in circuit-switched hardware and software pull the plug on their existing infrastructures and adopt Internet-protocol (IP) networks?

The answer depends on how big that investment is, and at what point a carrier can economically justify moving the legacy voice traffic--the roughly 45% of total network traffic that still generates 80% of revenues--from the circuit-switched network over to the more efficient IP network.

Is there a pain-free, technically efficient way to get from here to there? If you're an ISP, the answer is "yes." Simply deploy terabit switch in the core and move ATM switches from the core to the edge and you're there. But there is more to it for carriers. With substantial investments in TDM and curcuit switching equipment, a more delicate migration plan is needed. Carriers need to get past the current debate over which is better, optical switches or terabit switch routers. Both of these technologies can coexist as the primary tools used to pave a path to carriers' future networks.

Like it or not, optical-switching technology will play a major role in next-generation carrier networks and terabit switch/routers that incorporate optical switching will make it easier to get there. Long-term, terabit routers are the future. An overview of these technologies and the reasons they were developed can help network planners make the difficult but unavoidable decisions about how to pull it all off while staying competitive.

These dicey migration decisions stem from unbridled traffic growth. Internet use is growing fourfold annually with users demanding a broad array of access speeds and quality-of-service (QoS) levels. This exponential growth in bandwidth demand has led to rapid deployment of dense wavelength-division multiplexing (DWDM) technology in providers' network backbones.

Using DWDM to amplify single strands of fiber by an order of magnitude, carriers now have a cost-effective alternative to deploying more fiber. DWDM systems can turn standard 1310-nm or 15xx-nm lasers into a rainbow of light frequencies, creating 160 "virtual fibers" from one strand. A 10-Gbit/sec strand can carry 1.6 terabits of revenue-producing traffic--removing the need to deploy 160 individual 10-Gbit/sec fibers.

The capacity increases sparked by DWDM breakthroughs ignited debate over how to carve up and deliver differentiated services. Two technologies emerged which allow carriers to directly interface with the Internet's optical core, creating backbone devices that improve performance and reduce capital and operational costs. These backbone devices--optical switches and terabit routers--were designed to facilitate the delivery of high-margin, value-add services.

In fact, they compliment each other. Optical switches can wring more out of today's circuit-based networks, helping public-switched telephone networks (PSTNs) perform more efficient Layer 1 trunking. Terabit routers with integrated optical technology provide the critical task of access grooming with extremly high granularity and are further optimized to improve data handling with multiprotocol label switching (MPLS) and traffic engineering. Long-term, however, terabit routers are positioned to become the central-office switches of the new public network (see Figure). As circuit-switch technology is replaced by packet-based transmission, terabit switch routers will make a new communications platform possible from which providers can launch feature-rich, high-margin services with guaranteed QoS at prices affordable to the masses.Th Acf188

The above figure shows that as the majority of network traffic moves from voice to data, carriers can make their current circuit-based network more efficient by deploying optical switches and terabit switch routers into their core (left side of diagram). The long-term view of the carrier network (right side of diagram) emulates an ISP where there is no such thing as isochronous traffic and all devices are optimized for routing IP packets-the terabit switch/router becomes the sole central-office core switch of the new packet-based public network. Exponential growth of IP traffic is supported by turning up new DWDM lightwaves.

Optical-switch technology is designed to deliver higher margins from the circuit-based public network today. Terabit switch routing has two roles--to provide the required traffic grooming for optical switches in today's circuit-based network and to enable the creation of a new packet-based network capable of supporting innovative high-margin applications. Optical switches provide a more economical and efficient means of supporting real-time time-division multiplexing (TDM) traffic over an optical core. Terabit routers are optimized to support the explosive growth of IP data packets. Despite different development paths, optical switches and terabit routers can leverage one another to flourish in the carrier space.

The premise behind optical technology is circuit-based switching and TDM, both optimized for isochronous traffic. Based on TDM, or space-division multiplexing, connection-oriented circuit switching provides consistency and minimal delay by switching among fixed traffic channels every 125 microseconds, whether the circuit is running at a speed of 1.544 Mbits/sec or 10 Gbits/sec.

Data packets have grown nearly overnight to more than 50% of the aggregate network traffic. In response, vendors have developed enhancements to circuit switching to help the PSTN support data and voice traffic more efficiently, for example, the task-specific protocols: Asynchronous Transfer Mode (ATM) and Synchronous Optical Network/ Synchronous Digital Hierarchy (SONET/SDH). While both protocols improve performance, their addition to networks produced a redundant overlay with switching and multiplexing taking place at multiple layers. For example, IP and packet switching is performed at Layer 3, ATM cell switching occurs at Layer 2, SONET/SDH framing happens at Layer 1, and DWDM light frequencies are created at Layer 0.

Circuit switching also created a tangle of virtual circuits. A problem developed because every node needed a direct connection to every other node in the network. Something known as the n2 problem, where "n" number of nodes need n2 circuits in between them to provide connectivity across the network. This mesh of circuits, a result of network growth, has reached the point where it is unmanageable for service providers.

Optical switching can reduce the n2 problem at the physical layer. Today's optical switches are actually digital crossconnects with fiber interfaces that switch electronic signals, optical switches provide a more efficient means of switching tandem, statistical, or TDM traffic in a circuit-based backbone. They add intelligence to Layer 1 and, with the help of terabit routers, can eliminate two unnecessary technology layers from the core--ATM and SONET/SDH. Thus, optical switches offer a good solution to cut capital and operational costs.

Next-generation providers can deploy optical switches to replace SONET/SDH rings with a mesh network that performs more-efficient trunk switching and doesn't require the dedication of a substantial amount of bandwidth for network protection. Optical-switch vendors will most likely develop some kind of a dynamic link state protocol--analogous to open shortest path first (OSPF), private-network node interface (PNNI), or MPLS--that can provide Layer 1 dynamic rerouting and automatic protection in a mesh environment. Optical switches have drawbacks, however. They do not support access interfaces below 2.5 Gbits/sec. They work only at the wavelength level and are oblivious to the data inside a light channel. Thus, optical switches are unable to identify packets and offer differentiated services. They require packet-based devices such as terabit routers to perform local-access traffic grooming.

Optical switching would make sense as a solution to support access traffic grooming if the current fourfold network growth was due to only a few users requiring four times the access speed. Most users, however, are not requesting point-to-point, clear-channel access speeds on an OC-48 (2.5-Gbit/sec) basis. Every optical-switch port needs a terabit-router port, so that an OC-48 pipe can be carved up for many users in a point-to-point or point-to-multipoint configuration. Finally though, optical switches can improve the overlay phenomenon and, in conjunction with terabit routers, supplant ATM and SONET/SDH, the n2 problem and overlay networks will still exist to some extent as long as circuit switching technology remains present in the network.

Terabit technology is the basis for a parallel effort to create a new network optimized for the exponential growth of IP data traffic. As isochronous traffic (i.e., voice and video) evolves into compressed packets and is treated like data, many providers will welcome a chance for a fresh start in building their networks. Based on connectionless packet switching that does not require signaling, circuits, or every-node to every-other-node connectivity, terabit routers are optimized for IP.

In packet switching, data is segmented into fixed or variable packets prefixed with a header. The network transmits data in a store-and-forward manner according to the fields in the header, transferring packets across the infrastructure to their destinations. Terabit routers are versatile in that they route access traffic and switch tandem trunk traffic within special hardware application-specific integrated circuits capable of providing QoS at terabit speeds.

As a result, terabit routers can solve many of the problems commonly associated with circuit-based networks. These devices allow IP traffic to interface directly with DWDM systems, without requiring intermediate boxes. Terabit routers also eliminate the n2 problem at Layers 1, 2, and 3, because they are packet-based and do not require every network node to have a direct circuit connection to every other node.

As for services, terabit routers can isolate individual packets and make routing decisions based on the information bits inside. This means routing efficiency and the ability to provide service granularity down to the size of a single 40-byte packet, which is critical for supporting the massive growth expected in the number of public-network users. Providers need a technology that delivers higher operational efficiencies and can support customers' appetites for a variety of service levels and access requirements ranging from sub T-1 (1.554 Mbits/sec) to OC-192.

On the expectation that photons will be the foundation of the next-generation optical Internet, billions of dollars will be spent developing equipment to create, amplify, route, and switch wavelengths that carry a limitless supply of digital information. This spending will be driven by an explosion of data traffic, accelerated by the mass adoption of broadband access technologies such as cable modems and high-speed digital subscriber lines. Along the way, expect optical switches and terabit switch routers to help squeeze higher margins from today's older TDM-based public network.

Terabit switch routing alone, however, will be the basis for a new packet-based network designed to support a broad range of applications like e-commerce, multimedia, video-on-demand, videoconferencing, virtual private networks, multicast, and virtual gaming. By merging the features of task-specific devices, terabit routers will simplify core networks, cut operational costs, and improve performance, creating a network architecture with lower per-bit transport and ownership costs. The lingering question for carriers is, how quickly do you want to consolidate the boxes in your central office and transition to a network that is optimized for IP?

Stephen Duffy is a product-line manager at Avici Systems (North Billerica, MA).

More in Network Design