Evolving toward layered logical networks
A shared language created by emerging standards will drive next-generation layered logical management systems.
JEAN LAWLIS, Astral Point Communications
In today's telecommunications market, the ability to manage a growing and increasingly diverse set of services is a competitive necessity. Service providers must move beyond current management solutions, which offer flat views of physical topologies and network equipment.
The movement toward layered logical approaches is gathering momentum with the emergence of industry standards that are creating a shared language for describing network capabilities and a new framework for sharing information and network control. Together, the standards and framework will enable a new generation of management systems capable of rapidly adapting to changing technology, new service concepts, and business partnerships.
By providing for standard interfaces and common infrastructures, the emerging specifications enable "loose coupling," which is a key principle of layered logical management. With previous management approaches, applications needing to share information were cemented together with expensive, customized software that enabled them to interoperate-but at a price. Making changes was a slow, expensive process. The benefits of new technology were only slowly realized; and profitable new services frequently were delayed for years. In contrast, loose coupling provides the ability to introduce changes and rapidly evolve both the underlying technology and the management applications.
Network technologies logically divide into layers based on the functions they provide, but questions remain on the best way to deploy and layer these technologies. Should Internet Protocol (IP) routers be connected via protected SONET connections over DWDM? Should ATM networks run directly over DWDM networks? Because technology evolves so quickly, the methods that work today may not work tomorrow. But one thing that will not change is the service provider's need for management systems that can abstract the right view of the network for the problem at hand.
For example, DWDM equipment operates at the physical layer, providing a physical channel for data to flow (a wavelength of light over a fiber). These wavelengths are used to connect a variety of equipment with sophisticated packet routing and switching, protection switching, and error detection and correction capabilities. When viewed at the optical (DWDM) layer, this equipment may physically look like a mesh network. When viewed at the SONET layer, it may logically look like interconnected rings. These differences are more than academic. For instance, when planning for new fiber capacity, viewing the DWDM network and its current wavelength utilization is meaningful. But if there is a fiber cut, service providers will need a different view to find out which services are affected and how. Logical layering implies appropriate views of underlying network resources, but it also implies more.
The telecommunications-management network (TMN) hierarchy has been used as the basis for building telecommunications systems for 20 years. TMN provides a logical layering of functionality (see Figure 1). At the bottom of the TMN hierarchy is the physical network equipment and the native functions it needs to present to higher-level systems. Above that level is the element-management layer (EML), where element-management systems (EMSs) manage collections of network elements (NEs), monitoring their health, providing provisioning cap abilities, and storing data about NEs for analysis. The next layer is the network-management level; here the network-management systems (NMSs) take a network-wide view, create connections across the network, and analyze faults based on network behavior. Above that level are service- and business-management systems that handle customer and service information and functions such as creating and executing service orders or producing monthly bills. These systems are less concerned with the details of the underlying equipment. They do, however, need to interact with the underlying network through the lower-level systems to communicate accounting data collected by the NEs and create connections to support services. In TMN layering, every system receives only the data and control necessary for its tasks from the system below it in the hierarchy. Moving down the hierarchy, service-management information is abstracted out and moving up the hierarchy and network equipment information is abstracted out (see Figure 2).
TMN has served the telecom industry well, but changing technology is forcing management architectures to evolve and blurring the boundaries between TMN layers. One of the forces driving this evolution is the trend toward distributed systems taking on client/server roles. If a system has services to offer, it takes on the server role. If the same system is using services, it is a client of those services. The same system can perform multiple server roles and also be a client to multiple other servers. With the client/server model, when systems functioning at the business systems layer need the services of systems at the EML, the needed information no longer must go through two intervening systems. The business system can simply become a client to the EMS. Similarly, when two systems as the same level in the TMN hierarchy need to exchange information, the one providing the information becomes a server to the other (see Figure 3).
In the 1990s, architectures such as the Common Request Broker Architecture (CORBA) and tools/programming languages such as Java for distributed application support and eXtended Markup Language (XML) for platform-independent data exchange emerged. Pro viding an environment for distributed computing, these programming languages and architectures support a framework that makes logical layered applications possible. A generation of programmers has grown up with these programming paradigms, and work in standards bodies and universities continues to refine them. The tools for layering applications are now widely available, efficient, inexpensive, and improving every year.
The evolution of network management toward layered logical approaches will also eliminate data-management inefficiencies. Today, the same data frequently exists in five or more NMSs. Getting data from one system to another is a manual, expensive process, and inconsistencies can develop between the systems. For these reasons, the cost of data management is much higher than it should be. A way to reduce data-management costs is by recognizing that some data is naturally centralized, while other data is decentralized. Layered applications are supported by frameworks, which make data locations transparent and allow optimal data management.
The intelligence of network equipment effects layering, as well. In the 1970s, the only network intelligence was in the switching systems in central offices. Lacking intelligence, the transmission equipment connecting these systems and distributing traffic to homes and businesses could not be remotely queried or provisioned. All changes had to be made at the actual physical location. Intelligent NEs became available in the 1980s; however, NEs could only provide their own information and had little or no awareness of the network. As such, network-oriented operations required separate management systems.
This situation improved in the 1990s with the shared intelligence of SONET systems and the network intelligence inherent in ATM and IP. With these advances, functions such as setting up and tearing down connections or diagnosing and correcting problems increasingly were delegated to the network. Looking back to the TMN model, the EML and network-management-layer (NML) functions increasingly are pulled into the network itself, flattening the management hierarchy.
In the past, management systems had little or no need to communicate with systems beyond the service provider's boundary. Today, networks are increasingly shared. Every large carrier has become a carrier's carrier. All large business customer want more information and control over their telecom operations. That requires sharing of information not assumed in the original TMN model; service providers are beginning to provide their customers and business partners with layered network views in which only the customer's resources are visible and controllable.
Enabled by new layering paradigms, the past few years have seen a proliferation of independent software vendors (ISVs) that offer management systems. Some specialize in analyzing and creating reports on the performance of the network; others collect accounting data and produce bills; and still others create end-to-end services across multitechnology, multivendor networks. Successful ISVs establish partnerships with other ISVs, thus enabling their applications to communicate. For example, such partnerships may enable inventory-management systems from one vendor to adjust inventory levels appropriately when a service is created by a system from a different vendor. Today, service providers can select from a dozen or more ISVs in each application space. Service providers can then build complete operations environments from these applications; systems don't quite "plug and play," but integration is getting easier all the time. The whole process is an order of magnitude faster and less expensive than the homegrown or heavily customized operations environments of the past.
For the equipment vendors, a model has emerged that makes supporting logically layered applications simpler and less expensive. The vendor's EMS is designed to perform the lower-level services of TMN and must interface directly or indirectly with all the integrated applications above it. The multitechnology EML-to-NML interface being worked on at the TeleManagement Forum is emerging as a common interface, with broad functional support that vendors across the industry are using as a single interface to support all the needs of the applications layered above it. Mirroring this advantage, with a single interface that works for multiple vendors, the job of ISVs is simplified, as well.
Clearly, layered logical management provides the best approach for enabling multitechnology, multivendor networks to be viewed differently depending on the purpose of the application at hand. By giving service providers the ability to manage a growing and increasingly diverse set of services, layered logical networking approaches will almost certainly mean the difference between success and failure in the future.
Jean Lawlis is the manager of product management for network and service management systems at Astral Point Communications (Chelmsford, MA).