With bandwidth demands satisfied, network planners are now pressured to reduce network architecture costs. Carrier network profits are being driven by the cost of operating the network, and less by underlying capital and depreciation. Based on 1998 FCC data, nearly 20 percent of the industry's expenses are operational costs. With capital costs for equipment and cable dropping rapidly, operations costs will continue to grow in relation to overall expenses.
One potential area of savings lies in the operations support systems (OSS). The optical network can be provisioned and operated in the same way it has been for the last 30 years-perpetuating centralized, proprietary OSS systems-or it can move to a distributed model where the network is the database.
Wavelength and circuit-routed systems provide a radically simplified approach in that only source and destination information is required to deliver services, and the system itself routes the service. Services are provisioned rapidly and applied without requiring knowledge of the network topology.
Transport networks today are operated based on a combination of Sonet/SDH and mainframe computer systems. Sonet and SDH systems provide an automated platform upon which centralized, mainframe-based information systems are used to operate, administer, maintain and provision (OAM&P) services. Each network element knows its configuration and the state of each circuit that traverses its domain. These elements are automated but not intelligent, in that they have no knowledge of the configuration of the network.
Centralized computing systems are used to apply services across the automated transport network on a point-to-point, link-by-link, circuit-pack-by-circuit-pack basis. This requires the OSS to inventory and know the exact state of all objects-hardware and reconfigurable elements.
The transport element itself simply manipulates the circuits as directed by the centralized provisioning systems.
Basically, the OSS systems themselves are proprietary, expensive and slow to migrate to new technologies. Typically, it takes 18 to 24 months after the technology is available for the network for OSSes to be enabled, if they can be enabled at all.
Centralized OAM&P with automated network elements fits well with low-growth, low-innovation networks. Core network backbone growth of 100 to 300 per-cent per year is combined with disruptive-and accelerating-technology cycles. Therefore the fundamental approach to managing the network must change to realize substantial improvements in operations.
In the Internet age, centralized network-management systems and network intelligence will fade away as open, intelligent, network-aware systems emerge. Distributed networking, like distributed computing, will allow networks to scale up in capacity and not be limited by the inability of support systems-and technicians-to manage growth. With the management systems shifting toward the distributed model, the network elements themselves will have to undergo some fundamental change.
Wavelength routing technology is based on systems that are intelligent, network-aware and able to leverage optical internetworking in carrier networks. Wavelength routing elements use distributed intelligence in the system architecture and in the network to provide a platform for rapid service provisioning and restoration. There are direct, established communication paths between wavelength routing systems such that each element knows the configuration of the network and of its neighboring systems.
As network-aware systems, wavelength routing elements are not dependent on legacy OSS to provision and restore the network. Because the wavelength routing element is network-aware, it can route services through the network based only on source and destination routing. Therefore, to provision services, the network-not a centralized OSS-provides the intelligence and routing.
A service provisioning case, where the effects of combining of wavelength granularity and the application of routing techniques illustrates these points. In this model, four OC-48s are required between two points in a core optical backbone. The first analysis uses legacy STS-1 granularity and operations to provision these services; the second approach uses wavelength granularity with legacy operations; and the third approach uses both wavelength granularity and routing (wavelength routing).
The specific service model requires four OC-48s to traverse five hops across a core transport network.
In the STS-1 model, 1,200 objects must be manipulated to create the service. By moving to wavelength granularity alone, this reduces the number of objects to 72 because a higher level of granularity (bigger pipes) is being used to manipulate the network. Finally, by combining wavelength granularity with wavelength routing principles, this can be reduced to only two objects. Only two objects are required because in wavelength-routed networks, only source and destination information is necessary, and the systems themselves route the service and configure the hardware.