Together, automated provisioning and optical networks will open a whole new world of responsive customer-driven services.
MARTIN J. STEINMANN, Syndesis
I can't wait to pester my grandchildren with stories about the old days: "You kids don't know how lucky you are. When I was your age, bandwidth was a scarce commodity. It took hours to download new computer games or music files. When you tried to watch a movie over the Internet, it looked awful-and even if you could afford more bandwidth, you had to wait months to get it.
"Then, optical networks and automated provisioning came along, and everything changed. Now, you can watch a movie or talk with your friends over the Internet and the pictures are perfect. You can download a new game or MP3 in an instant. And if you need more bandwidth-although I can't imagine why-you just click a button and presto...more bandwidth. Children today are so spoiled."
No doubt about it, optical networks will change everything. DWDM is already erasing long-haul bottlenecks and will soon do the same in regional and metropolitan area networks (MANs). And once optical switches are deployed, carriers will be able to concatenate point-to-point wavelengths, or "lambdas," into high-capacity wave paths that span the public infrastructure.
But that's only the beginning. The real payoff comes when wave paths can be provisioned automatically. Then, savvy service providers will use transport capacity and just-in-time provisioning as differentiating elements. The wait for service activation will drop from weeks to minutes. Lambda-based traffic segregation will solve the quality-of-service (QoS) problem, and struggling applications like voice-over-Internet Protocol (VoIP) will finally become practical.
Before this vision can be realized, however, networks need to slim down. Today's typical wide area network (WAN) carries IP on top of ATM, ATM on top of SONET/SDH, and SONET/SDH on top of DWDM (see Figure 1). When first conceived, this layering made sense. But as IP and DWDM evolve, the ATM and SONET/SDH layers are becoming superfluous. Worse yet, they are getting in the way of service improvements that customers want and carriers are eager to provide.
SONET/SDH circuits, for example, take forever to provision. First, a network engineer must review current asset inventory and channel assignments to locate available capacity. Then network operators must manually configure every crossconnect switch and add/drop multiplexer between the end points. Even using the Bellcore (now Telcordia)-developed Trunks Integrated Records Keeping System to track inventory and facilitate workflow, it still takes 45 to 60 days to set up a new coast-to-coast circuit.
To add insult to injury, network complexity and the rigid SONET/SDH bandwidth hierarchy inhibits scalability. With DWDM, network operators can increase point-to-point bandwidth merely by lighting up another wavelength on an existing fiber. But to increase capacity at the SONET/SDH layer, every SONET/SDH device on the fiber must be upgraded-an expensive and disruptive proposition. Also, since SONET/SDH supports only a few interface speeds-OC-3 (155 Mbits/sec), OC-12 (622 Mbits/sec), OC-48 (2.4 Gbits/sec), etc.-it's often impossible to match bandwidth costs with bandwidth requirements.
The solution, many believe, is to layer IP directly over the optical substrate. Then service providers can allocate wave paths on demand and tailor them to fit specific IP-based services. Optical transport can become an integral part of IP service delivery.
To bring IP and DWDM together, new capabilities must be added to both. A framing standard is needed for carrying packets directly over lambdas. Signaling standards are needed so that IP devices can control optical resources. Industry organizations like the Optical Internetworking Forum and the Optical Domain Service Interconnect (ODSI) are rushing to meet these needs. An interface under development by the ODSI, for example, lets switches, routers, and optical edge devices request connections across an optical backbone (see Figure 2). Requests can include criteria such as low latency or fast recovery from fiber cuts. If the criteria can be met, the new bandwidth is created in seconds. Another effort-dubbed Multiprotocol Lamb da Switching (MPlS) -extends Multiprotocol Label Switching (MPLS) so routers can cooperate as peers with optical core switches for dynamic creation of end-to-end services.
Layering IP directly over DWDM promises significant benefits to service providers:
- Equipment costs will drop since fewer boxes will be needed.
- Network operations will be simpler since there are fewer layers and fewer boxes to manage.
- Bandwidth efficiency will increase since fewer protocol layers mean fewer headers and less packet fragmentation.
- Networks will become more scalable since bandwidth can be added without equipment upgrades.
More importantly, with ATM and SONET/SDH out of the way, automated provisioning systems will gain direct access to DWDM resources, and dynamic wave-path provisioning will become a strategic weapon for network service providers. Providers with the latest optical gear and the most advanced provisioning systems will jump light-years ahead in the race to attract and hold customers.
Dynamic wave-path provisioning will let service providers light up new connections faster than their competitors. Activation times will drop from weeks to minutes. Providers will be able to meet short-duration requirements-for special broadcasts, videoconferences, or third-shift backups-responsively and economically. With dynamic wave-path provisioning, providers can even offer services that automatically add and remove bandwidth as demand varies.
Unconstrained by the SONET/SDH hierarchy, service providers can use DWDM lambdas to furnish predictable, guaranteed transport for any service, from a simple T1 (1.5-Mbit/sec) link to a high-capacity Gigabit Ethernet connection. Flexible multiplexing schemes will ensure that lambdas are used cost-effectively, while the bit-level transparency of optical transport will reduce the need for protocol conversion.
The combination of IP over DWDM and automated provisioning will enable carriers to offer new services and service-level agreements (SLAs) with confidence. For example, today's voice-over-IP technology simply does not work. IP cannot deliver the necessary QoS, and SONET/SDH does nothing to help. But with techniques like MPlS, carriers can dedicate whole wavelengths to VoIP traffic and satisfy even the most stringent QoS criteria. In a similar manner, real-time video, virtual private lines, secure virtual private networks, and differentiated Internet access can all be delivered with guaranteed QoS and ironclad SLAs.
The effect on the WAN industry will be revolutionary. Not only will two major technologies-ATM and SONET/
SDH-decline in importance, but control over network resources will shift dramatically. End users will no longer be at the mercy of their service providers, waiting months for every new service and settling for mediocre QoS. Instead, users and their applications will interact with the network to get the services they want, when they want them.
As compelling as the IP-over-DWDM vision is-streamlined operation, bandwidth on demand, guaranteed QoS, customer self-management-none of it works without automated provisioning. Without automated provisioning, service activation will still be slow and unreliable. Network operators will still have to configure every switch and router, one by one, specifying ports, allocating lambdas, assigning addresses, and so forth. Manual provisioning will remain a stumbling block and the full potential of IP over DWDM will never be realized.
With automated provisioning in place, everything will be different. Network providers will define services in advance. When a customer requests a new service, an operator will merely select the appropriate service profile and add customer-specific information such as incoming port number or bandwidth level. The provisioning system will translate the profile into configuration commands for each device in the service path and send the commands to the network. In minutes, the service will be up and running, giving the provider an instant advantage over competitors with manual systems.
Even when IP devices can signal for wave paths on their own, there will still be a need for automated provisioning. MPlS, for instance, uses open shortest path first or private-network node interface to layout paths through the network. When a router notices congestion between two points, it can tell the optical core to add another lambda to that path. When the burst ends, the router can return the lambda to the resource pool. To meet the service provider's business goals, this process must comply with policies governing bandwidth costs, QoS, protection and restoration, load balancing, explicit routing, backup routes, prioritization, preemption, constraint-based routing, diverse routing, etc. Manual administration of these policies will slow down service activation and drive up operation costs. An advanced provisioning system will capture these policies, imbed them in network elements, and apply them automatically as edge devices-or network operators-signal for more lambdas.
Many network equipment manufacturers offer element- or domain-network management for their own products. But a typical IP-over-DWDM infrastructure includes devices from several suppliers-DWDM gear from one company, routers from another company, DSL boxes from a third source, etc. To create an end-to-end service, the network operator has to match up port assignments, coordinate virtual circuit identifiers, synchronize bandwidth levels, and so forth, between different manufacturer's devices-a slow, expensive, and error-prone process. To yield maximum benefit, therefore, an automated provisioning system must be complete, reaching every device in the service path and including optical resources as an essential ingredient.
With a provisioning system that spans multiple technologies and vendors, a single touch activates a complete, end-to-end service instantaneously and error-free. Also, by imposing a layer of abstraction between service profiles and specific technologies, the provisioning system enables each service provider to offer consistent services throughout its infrastructure and maintain that consistency as the network evolves.
To enable integration with other parts of the carrier's operations support system, the provisioning system should also include a bidirectional flow-through interface (FTI). Using the FTI, order-entry and workflow applications can trigger service activation electronically, eliminating paperwork and further reducing provisioning delay (see Figure 3). The FTI can also link multiple carriers so that a long-haul provider, for example, can automatically signal a local provider for an access circuit.
By adding a Web interface to its order-entry system, a service provider can even open its network to customer self-provisioning. Based on predetermined policies, customers can order new services, increase bandwidth allocation, or modify QoS parameters on their own. Their requests pass automatically from the order-entry system to the service activation system with zero paperwork and zero involvement by network operators. In minutes, ports are configured, lambdas allocated, priorities set, and the service activated-the ultimate in responsive provisioning and low-cost operation.
Automated provisioning is most effective when backed by advanced in-service asset management. An IP-over-DWDM network may be simpler than its predecessors, but keeping track of assets such as switch ports and lambdas is still a challenge. Manual inventory procedures still result in stranded resources and broken connections. To guarantee rapid, accurate service activation, the provisioning system must use the network itself as the database of in-service assets. It must interrogate network devices directly to learn which resources-trunks, ports, lambdas, output buffers, address ranges-are in use and which are available.
With the network as the database, the service provider can accurately assess resource availability before accepting a new service order. By including lambdas in the resource pool, the provider can make transport an essential part of each service. The provisioning system can check, for example, the availability of unused wavelengths and IP address blocks before turning on a broadband Internet service. Without this capability, manual procedures will continue to waste resources, retard service activation, and compromise SLAs. With this capability, innovative providers can use advanced asset management to create a variety of new revenue-enhancing services as well as identify and resolve problems before they impact service delivery.
Moreover, automated real-time asset management will help service providers get the most out of their transport networks. Even though DWDM adds tremendous capacity to the wide-area infrastructure, many experts believe that networks will still be congested. By maintaining up-to-the-millisecond records of bandwidth utilization, real-time asset management will ensure that each wavelength is used as efficiently as possible without the agonizing delay of manual procedures.
With the network as the database, the provisioning system can also expedite bulk moves and other routine maintenance procedures. Consider a DSL-based differentiated Internet access service. Each time the service provider adds a backbone router to boost network capacity, hundreds of static routes must be deleted, then recreated over new links to re-balance the traffic with the proper QoS assignments. Manual "delete/create" cycles can eat up hundreds of operator-hours and cause unacceptable outages for hundreds of customers. By automating the process, a provider can slash operational costs and reduce service interruptions to just a few seconds. Moreover, with the network as the database, the maintenance application can verify that all the moves will succeed before changing any route tables.
WANs are at a turning point. Data is taking over from voice, packets are replacing circuits, and DWDM is ending the bandwidth shortage. Soon, streamlined IP-over-DWDM networks will replace cumbersome multilayer infrastructures, and DWDM lambdas will become strategic assets. Service providers who use these assets wisely will be richly rewarded.
Automated provisioning will be the key to success in this new era. The improvement in network services will be profound-and my grandchildren will take it all for granted.
Martin J. Steinmann is vice president of marketing at Syndesis (Toronto).