Think scale, MAN!
Optical packet networking can dynamically balance network scalability by providing a wide range of services.
DAN PORANSKI, Tropic Networks
During the salad days of the telecommunications industry, life was easy and predictable. For customers of telephone companies, the only service required was a voice conversation. Conversations were easy to model, circuits were easy to establish, and network traffic was very predictable. Probably the only way to experience this bygone time today is by taking a trip to the Smithsonian Institute.
Today, Internet and corporate data traffic has proliferated through the telecom infrastructure and far surpassed that of voice (see Figure 1). Unfortunately, the existing MANs were built using architectures and technologies geared toward predictable, low-bandwidth voice traffic and are not suitable for the demands of today's, and certainly not tomorrow's, data-centric world.
Carriers need to offer a much broader range of services to meet the communications needs of today's diverse customer base. However, MANs are losing the battle to keep up with the growing demands of data because current technologies such as SONET/SDH do not provide the scalability or the dynamic control needed to tune the network to best serve the needs of the wide range of necessary services.
Imagine this scenario. What if Avis, the car rental company, only had high-end luxury sedans by Mercedes Benz to rent? The business model would work fine if Avis had a large enough, and growing, customer base that sought the value in renting a high-end luxury sedan. But what if Avis was forced to rent that high-end automobile to all its customers? And worse, what if they could only charge what those intermediate, economy, and compact car customers were willing to pay? It's easy to see that the business model would quickly fail in that situation.
But that's exactly what is happening with MANs today! The use of SONET/SDH across the MAN to support every type of service simply does not scale. Carriers are offering a wide range of services but are forced to use the same high-end network resources to support everything from economy class to luxury class. Just as Avis tunes its automobile offerings to meet consumer demands, carriers need to tune network resources to most cost-effectively support the type of service being sold.
The clearest manifestation of the scaling problem with current architectures in the MAN is the tremendous amount of optical-electrical-optical (OEO) conversion being done. Today's MANs are paying a costly and needless OEO "tax" imposed on network services by existing network technologies. In addition to requiring that expensive SONET/SDH resources be provisioned across the whole service range, the lack of integration between electrical- and optical-switching elements-and the resulting inefficient use of space and power, among other things-compounds the problem.
The result is that network costs do not scale linearly as the network grows. Carriers are forced to overbuild the network to meet the most critical service-level demands. Eliminating the nasty OEO tax requires new solutions for dynamic control and scalability in the network. Bringing the capabilities of dynamic control and scalability to the MAN can be a daunting and expensive task for telecom carriers. But the reality of the situation is that carriers must adapt their metro networks or face dire economic consequences.
Optical packet-networking (OPN) systems are one of the emerging solutions for the MAN that can help carriers achieve the control and scalability necessary to reach profitability with a wide range of services. The vision of OPNs is simple: IP is the dominant services protocol, Ethernet is the preferred data interface, and optical wavelengths are the ubiquitous transmission medium. By integrating data services with optical transport within this vision, new levels of scalability and flexibility are brought to MANs that not only address the fast-growing demands of data, but also provide new benefits to traditional types of circuit traffic such as voice.
A good solution starts with a good foundation. For networking systems, a good foundation is embodied in an architecture that possesses four important attributes to enable new dimensions of scalability:
- Independence of physical topology.
- Allowing nodes to contain both electrical- and optical-switching and/or multiplexing functions.
- Enabling electrical and optical domains to scale independently.
- Providing unified management of traffic flows between the optical and electrical domains.
Much debate has revolved around which topology-ring or mesh-is best suited for applications in the MAN. Rings provide an acceptable topology solution for restoration, resiliency, and protection. But when traffic demands at a particular node increase, the entire ring must be upgraded to support the increase of what may only be localized traffic. IP traffic, on the other hand, is better suited for a mesh topology. However, scaling and congestion problems can occur at any interconnect in the middle of very large meshed networks. So which is correct for an optical packet network? The answer is neither and both.
Employing a physical ring-logical mesh topology helps improve connectivity with each wavelength that's lit. One of the benefits of this method is it eliminates the need to "stack" multiple unconnected rings on top of one another to achieve higher bandwidths. Another benefit is that the impact of any one physical fiber failure is reduced.
On the other hand, a physical mesh-logical ring topology allows the network planner to establish a separate path through the network for each lightwave. In large networks, this approach alleviates the problem of wavelength blocking. Topology independence is an important dial that the network planner can use to dynamically tune the infrastructure to best support the needs of particular services.
For example, if there is a large amount of traffic on the network that can benefit from one topology versus another, the proper choice for the current state can be deployed at the proper cost point. But if traffic patterns over time indicate that a different topology is better suited, the network engineer can flexibly tune to the new topology with minimal impact on services and additional cost.
The next step in building the foundation is the choice of building materials. Both electrical (packet) and optical (wavelength) switching elements are necessary to enable service delivery. The electrical-switching domain is a shared-network resource that can rapidly allocate bandwidth to new customers and existing customers who need more.
Switching and routing of packet traffic is very flexible because the sources and destinations can be any reachable node in the network. Also, while the speed of provisioning and flexibility of traffic are very real benefits, using pure electrical switching is very cost ly for transit traffic in large networks. There are high capital costs for transponders and large switching fabrics as well as high operating costs-power, cooling, and space-that put tremendous strains on carrier business cases.
In the optical do main, a wavelength plan can be created and fixed optical filters installed at strategic node locations to provide optical add/drop and pass-through capabilities corresponding to traffic flows. That's the least costly solution for transit traffic, but also the least flexible since fixed-wavelength devices must be physically installed to implement the interconnect topology. Additionally, wavelength planning exercises and truck rolls to provision services require a much larger piece of the operational expense budget than the shared-resource packet-switched network.
OPN systems provide both electrical- and optical-switching elements in a single package under a common management and control plane. If the two domains are independently scalable, network designers and operators can further tune network resources to best meet service requirements by easily shifting traffic between electrical and optical domains.
Technology-independent traffic engineering techniques using Generalized MPLS can provide packet aggregation using flow equivalency classes (FECs) to ensure the most efficient use of the network. Identifying candidate aggregate flows from the packet domain allows them to be "carved off" into the less costly optical domain as early as possible once they begin to transverse the network. That avoids the unnecessary power and cost associated with OEO conversions and electrical switching for flows that will transit intermediate nodes.
The scalability provided by an OPN allows multiple services to be offered on the same infrastructure. Not only can services be offered at different internetworking levels, but also at the appropriate level of availability (see Figure 2). For the highest quality of services, such as voice or mission-critical data, network resources can be dedicated and reserved to deal with failure scenarios. Low-cost services can take advantage of spare bandwidth in the network.
Scaling the service offering to achieve multiple cost-for-per formance points on a common infrastructure will minimize or eliminate the OEO tax. But that can only be realized when the network operators have multiple "dials" they can use to tune network capabilities that best meet their service requirements.
The benefits in achieving a dynamic balance with network scalability are significant. Carriers can now offer the broad range of services necessary to capture a diverse customer base. More importantly, they can now control the balance under changing network conditions, including rapid growth, and apply the correct network resources to a particular service.
With more dials to tune, they are better able to manage the balance between capital costs, operating expenses, and revenue growth. It all boils down to compelling and sustainable business cases for the services they provide.
Dan Poranski is vice president of marketing at Tropic Networks (Kanata, Ontario). He can be reached at the company's Website, www.tropicnetworks.com.