Freeing The "Gordian Knot" In The Metro.

Legends say that Alexander the Great came upon an interesting challenge when arriving at Gordia; a complicated knot that held a chariot together without any ends exposed. Metro network planners have a similar challenge when trying to cope with service provisioning requirements and the uncertainty of demands. Trying to forecast the timing, place, and size of traffic demands from different types of interfaces, protocols, and services is like guessing which beam will untie the knot. It is very difficult to provide services rapidly. There are many uncertainties, such as: When new emerging WAN Ethernet interfaces are likely to take off? Is the vision of Intelligent Optical Networks for long-haul networks mature enough and how should Metro services be handed off to it? What is the future of ATM? What is the time frame to move (if at all) to "all IP services?" All in all, the main difficulty is the multi service nature of each node in the Metro. Today, service providers use a myriad of technologies for collecting an d transporting these services from the network edge to the metro hubs and Metro networks end up looking like complicated knots. There have been several attempts to untie these knots, including pulling the SONET string, pulling the ATM string, pulling the IP string, and pulling several strings together.

Untying The Knot By 'Pulling The SONET String'

Still the most popular way to deal with the knot problem is by solving it with SONET or derivatives of SONET, which include time slot concatenation. The idea is simple--bring all the services to a SONET format by mapping them to SONET time slots. The basic problem with this approach is that the service providers still have the inherent problems of painfully slow service provisioning and data services that come with granularities not fitting SONET time slots. New generation SONET improves the granularity, but STS-l or VT-1.5 concatenation still cannot compete with packet-based solutions. Aggregation in the hubs is done by traditional TDM grooming and multiplexing. While it is a very efficient way to connect to voice switches, it is not an efficient method for connecting to IP routers. In addition, classical SONET platforms are not able to provide best effort and differentiated data services and need an additional envelope of equipment to support them.

Untying The Knot By 'Pulling The ATM String'

A few years ago, ATM was the textbook answer. Standards were developed to convert TDM and Ethernet to ATM. ATM edge multiplexers were installed in the edge and many service providers invested in ATM core switches. Restoration schemes, such as ATM VP rings, were developed as a substitute to SONET. However, service providers still wanted their TDM SONET cross-connect and tended to shy away from the ATM VP rings that could not support OC3, OC12, and OC48 channelized TDM. Even data oriented service providers demand good support for their TDM voice and expect SONET quality. Moreover, as GbE rates and fast Ethernet emerge, the conversion, translation, and concentration of IP/Ethernet to ATM and back to IP does not make much sense.

Untying The Knot By 'Pulling The IP String'

If solving the problem with ATM was considered to be the textbook answer of the past, solving it with IP could be considered the textbook answer of the remote future. Customers want solutions now and they want it provisioned fast. Currently, Voice over IP is not able to provide the resilient and robust service expected from SONET. While Layer 3 and above processing is now a strong trend in silicon implementations, the speeds that are required and the "multi service nature" of the problem leaves a lot of work to be done.

Untying The Knot By 'Pulling Several Strings Together'

Since the above approaches were not able to solve the problem, the next step pulls several strings together. In effect, this is what the service providers did simply because they did not have any better way until now. Metro access and metro core are built from separate technologies: installation and management of complicated smart edges. Interface cards take only specific services. Each time a new type of service is introduced, new switch modules and new interface cards are required. End to end management is performed separately for ATM, IP, and SONET and is primarily performed on a node-by-node basis. In turn, this led to the development of the "Multi Service Provisioning Platforms" and to the "God boxes". These boxes usually have few fabrics and provide smart, high layer processing--layer 3 and above. Moreover, adding a new service (i.e., Fibre Channel) requires yet another conversion that may not be standard or easy. In perspective, these boxes are conceptually far from what transport networks are.

Freeing The Knot-Native Form Service Centric Approach

Alexander the Great found a creative way to solve his problem; he hacked the knot with his sword! For service providers, freeing the Metro knot requires a new approach to the problem. This new approach is called the Native Form Service Centric approach. Instead of trying to guess which technology to convert to and pay the penalties of processing, fabrics, and uncertainty, it is more economical to efficiently collect the services in the edge in their layer 1-2 native form. Thus, eliminating the high layer processing on the one hand and using a single fabric that serves all traffic types on the other hand. This approach eliminates the high layer processing as there is no economical sense in all these conversions and concentration; service providers want as much transparency and reduced inventory as they can get. This approach uses single fabric because service providers want simplicity, reduced space, and power per lambda. The addition of another service does not require any special conversion. It requires only layer 1-2 mapping. The figure illustrates the benefits of this Native Form Service approach.

MPLS As A Technology Enabler

The "sword" required to hack the knot is the technology that enables this unified multi-service layer 1-2 collection and transport. The technology should allow for flexible mapping of the services on wavelengths, differentiation between delay-sensitive services (i.e., TDM voice, Storage Area Networks), and delay-non-sensitive services that run on the same wavelength. It should also allow for: fast forwarding and routing engine; support for current and future restoration schemes (SONET, Network wide Light Path restoration) and rapid end-to-end provisioning. A clear candidate is the Multi Protocol Label Switching (MPLS) technology, which is currently under advanced standardization process. MPLS integrates all these features when supported by the right all-packet fabric. With MPLS, each Metro node performs packet-by-packet label switch routing on an infrastructure of DWDM lambda connections between the nodes.

As the legends say, "the person who frees the knot will rule all Asia!" Indeed, service providers who elect to use the transparent native form service centric approach solution will rule their Metro. Forecasting demands and rapid service provisioning will become easier. The legend also says that King Midas, who turned everything he touched into gold, was the one to tie the "Gordian Knot". Can it be that service providers will now be able to turn their fiber into gold?

Hagay Katz is co-founder and director of product management and Dr. Michael Mesh is co-founder and CTO of PacketLight Networks (KfarSaba, Israel).