June 2000

BY RALPH PARKER

Networks around the globe are on a collision course. New network architectures are being rapidly deployed and are increasingly interconnected with other networks. Yet there is growing concern whether all these new networks will deliver their promised economic and social benefits or be plagued by interoperability problems. What is going to happen as a host of network protocols collide in the interconnected digital sea of worldwide communications and information transport?

A NEW DEVELOPMENT AND DEPLOYMENT MODEL
The telecommunications industry has had more than 100 years to perfect circuit-switched networks. Users now enjoy high-quality service even when their calls traverse multiple interconnected circuit-switched networks. Quality was achieved through an orderly and lengthy service development and deployment process for circuit-switched networks. For example, in the U.S., AT&T turned up the first digital T1 circuits in the mid-1960s, deployed digital tandem switches in the 1970s, digital end offices in the 1980s, and announced a fully digital network in the early 1990s.

Now, a more chaotic development and deployment model has emerged. The sheer number of carriers, vendors, and available technologies has changed the way in which network architectures are designed and deployed, and how services are offered to the consumer.

While large numbers of dedicated professionals work through the various standards bodies to create technical standards, development of network architectures by vendors and deployment by carriers is becoming less standardized with the imminent extinction of telephone monopolies around the globe.

Given the shift from orderly deployments under the regulated monopoly/oligopoly structure to the more chaotic deployment model of the new telecommunications world, how can the industry assure successful deployment of new architectures while maintaining service excellence? What steps might the industry take to make sure that emerging network architectures function as reliably as the Public Switched Telephone Network (PSTN)? Much of the answer lies in how well service providers and their vendors plan, implement, and manage network turn-up and internetwork interconnection.

One proven approach is to replicate new architectures in a laboratory environment and conduct end-to-end tests in advance of deployment. At Sprint, where deployment of the Sprint Integrated, On-Demand Network (ION) is underway, ION services are being rigorously tested in Sprint's test lab in Westwood, KS.

"To ensure quality in ION deployment, Sprint performs two-stage testing starting with extensive interoperability testing in Sprint's ION labs and then moving to field alpha and beta testing. Sprint's ION labs are designed to replicate the production of telephony and data networks from end to end including switching, signaling, and transport facilities of both local and long-distance networks," reports Tom Moore, director of ION integration testing.

Governmental action can drive the approach, in some cases for an entire country. A recent example is Argentina, and the Wall Street Journal reported on March 1, 2000 that "Argentina will attempt to secure $5 billion in fresh telecommunications investments over the next 18 months through tough deregulation policies that are part of an overall strategy to orient its industrial base toward highly skilled labor."

"Regulators also get involved. The FCC chartered its fifth Network Reliability and Interoperability Council in March to "evaluate and report on the reliability of public telecommunications network services in the United States, including the reliability of packet-switched networks."

Additionally, a multitude of industry fora and standards bodies continue to work on a range of standards and interoperability efforts. This includes Committee T1, ETSI, IMTC/iNOW, the Packet Cable Group, IN Forum, ATIS's Network Testing Committee, and others. ETSI for example, conducts "bake-offs" where engineers get together to test implementations against each other for the purpose of debugging standards and implementations at an early stage. Bake-offs typically occur when a standard or product is being developed. When standards are firmed up, efforts tend to shift to the vendors as they concentrate on bringing products to market. As decisions are made by carriers to deploy new architectures, they typically work exclusively with their own vendors and integrators.

A METHODICAL DEPLOYMENT APPROACH
Given the enormity of the deployment task, how is the industry going to deliver futuristic new services, work at Internet speed instead of telephone time, satisfy the regulators, and delight the shareholders -- all while maintaining service quality and reliability? It is often useful to break such a large complex undertaking into more digestible parts to attempt a solution. Reliable deployment of next-generation network architectures, including integration with the PSTN, might be approached through a three-step process that includes: Deployment planning, analytical examination of issues, and integration testing.

Step One
The first suggested step is deployment planning. A good plan lays out a comprehensive strategy with detailed steps for deployment. Besides program management, the plan should address capacity, performance, security, interoperability with other environments (such as operational and customer care systems and other networks), and reliability. Deployment plans with a complete and comprehensive strategy for reliability are likely to result in a superior deployment to one where a lesser reliability component is adopted by the service provider. A plan with a strong emphasis on reliability should exhaustively address capacity, performance, and interoperability. Business considerations may impose constraints on reliability planning, or could even result in going too far in the opposite direction. Engineering and operational groups should work closely with the strategic and marketing departments to make sure that the deployment plan is consistent with company and marketing strategies. Planners should also seek input from customers.

Step Two
The second step is an analytical examination of deployment issues. Integration of elements within a network needs to be well understood, along with interconnections to other networks. This work sets the stage for conducting tests during the third step. Again, the business philosophy and resources of the carrier are a factor in how exhaustive the analysis will be. Many carriers will contract with a systems integrator/consultant to perform some or all of the pre-deployment analytical work. These firms usually have established processes for such engagements, and a select few offer powerful network modeling and analysis tools.
The bulk of the analysis step will consist of assessing capability, capacity, performance, security, and reliability. The engineering and operational analysts should work closely with strategic and market planners to match the technical assessments with the planned footprint, scope, and marketing of the network being deployed. Modeling capacity and traffic growth over an appropriate planning horizon is an essential part of assessing reliability, because reliability should be evaluated over time as the network expands. Experience has shown that most network elements do not necessarily scale linearly, necessitating upgrades of some elements before others as the network expands.

Analysis will, by necessity, address intra-network interoperability and reliability, and should also cover inter-network interoperability and reliability. During the analysis step, requirements for each should be identified and investigated. Each carrier will have to decide the point at which they will be contacting potential carriers to negotiate interconnection. Technical, operational, business, and regulatory considerations weigh into this important decision as to when to initiate contact, with whom, and for what types of interconnection.

Step Three
The third step is integration testing. Comprehensive testing of network elements and related systems prior to deployment is essential in ensuring reliability. Where tests identify problems, corrective actions must be taken and tests repeated to confirm fixes. Integration testing is most efficiently accomplished through a methodical approach, which should include:

• Establishment of objectives and plans with corresponding budgets and resources. Planning is very important in achieving maximum test coverage with the fewest test cases. Include contingencies for Murphy's law -- things will go wrong during testing!
• Element conformance testing. Each element should be tested against applicable standards or specifications to assure that the element functions as intended.
• Element-to-element interoperability testing. By replicating the network architecture to be deployed in a laboratory environment, interoperability among the different elements can be tested. For example, interoperability testing between a user gateway and an announcement server.
• End-to-end testing. The network should be tested "end to end" from the user's perspective. Drawing from the planning and analysis steps, all relevant permutations of user services and network configurations should be tested. Variables may be introduced to simulate real-world operating conditions such as high traffic loads, element and link failures, delay, etc. Security also needs to be tested.
• Systems integration testing. Operations, network management, customer care, and billing systems will need to operate within the architecture being deployed. Reliability is as dependent on operating processes and systems as the network elements themselves. Support systems should be part of the lab testbed to enable testing in a simulated production environment.
• Interconnection conformance testing. If the network to be deployed will be interconnecting with other networks and architectures, requirements of the interconnecting carriers (plus any regulatory or government requirements) will determine what testing is required. Interconnection is likely to include, at a minimum, providing other carriers with interface conformance documentation. In the U.S., most incumbent carriers have specific requirements that must be met in order to interconnect. Typically, carriers or vendors conduct or contract for conformance testing and go through a certification process with the interconnecting carrier.
• End-to-end interconnection testing. Reliability is often assumed on the basis that if everything works as designed within a network, and the internetwork interface conforms with applicable standards and specifications, end-to-end reliability across interconnection is assured. While this has often been the case for circuit-switched networks, it's still a major question mark for packet and hybrid packet/circuit-switched networks. Views often vary among vendors and carriers depending on their business philosophies and experiences and the economic, political, and regulatory climate.

A CALL FOR ACTION
Service quality and reliability are mission critical for the user community. From fueling commerce to assuring priority services for national security and emergency response organizations, telecommunications service is a vital resource for users around the globe.

"The continued reliability of public networks, which are already becoming a hybrid of packet and circuit-switched technologies, is being closely watched by those of us responsible for providing reliable communications for federal, state, and local agencies, says John Graves, program director for Government Emergency Telecommunications Service (GETS). "We're looking for confirmation from the industry that reliability will be maintained while at the same time next-generation networks will provide powerful enhanced capabilities for the National Security and Emergency Preparedness community."

Users, carriers, vendors, shareholders, and regulators each have a stake in how effectively the global telecommunications industry deploys and interconnects emerging new network architectures. The three-step methodical approach -- planning, analysis, and interoperability testing -- offers carriers a way to step up and meet the challenges of deploying new network architectures. Vigorous and vigilant intra-network testing by carriers and their vendors is key to assuring intra-network reliability.

Carriers may also want to look at industry fora as viable venues to address inter-network performance and reliability. Bona fide, multi-network, multi-architecture, interconnection tests among emerging packet network providers and PSTN carriers would provide the industry a means to objectively assess reliability of interconnected networks. It may make sense, within the applicable laws and regulations, for carriers and vendors tackling similar interconnection and reliability challenges, to work together. There have been numerous instances where competitors have found common ground on which they can work cooperatively, and later compete vigorously for customers.

Ralph Parker is a senior consultant in Telcordia's Network and Operations Integration Practice. Telcordia's Network and Operations Integration Practice provides integration testing and consulting services to carriers, vendors, industry groups, enterprises, and the public sector worldwide. Ralph can be reached at [email protected]. For more information, visit Telcordia's Web site at www.telcordia.com.

BY FRED ENGEL

Due to overworked and understaffed IT departments, today's corporations are increasingly relying on service providers to maintain their networks and applications. These service providers include traditional carriers, managed network service providers, Internet service providers, Web hosters, and application service providers (ASPs). In each case, it is imperative that both parties (customer and service provider) be able to measure service delivery against objective goals for availability, responsiveness, and overall quality of services. This is typically accomplished by defining a Service Level Agreement (SLA). From the service provider perspective, customer satisfaction and loyalty can hinge on meeting these expectations -- and financial penalties can result from missing them. This article describes how SLAs should be defined, how they can be managed, and how they can be used as a valuable planning and marketing tool.

SERVICE LEVEL MANAGEMENT
The key benefit of an SLA is that it sets realistic expectations for both the service provider and the customer. While in some cases the SLA is a formal agreement with financial penalties attached to non-compliance, in other cases it may be an informal understanding. In either case, the SLA creates a need for effective service level management. Service delivery problems must be quickly detected and remedial actions taken. Downtime or delay in service delivery can mean millions of dollars in lost productivity and revenue -- for the service provider and its customers. Typical service level management involves a continuous process of evaluating and reevaluating business requirements, translating them into performance metrics, monitoring the results for compliance, and taking remedial actions if necessary.

COMMON PERFORMANCE METRICS
To successfully manage an SLA, objective verifiable performance metrics must be used. At the same time, a monitoring and reporting system must be in place to ensure that the metrics can be measured accurately.

For example, a service provider may offer VoIP and Web hosting services to their customers. In VoIP, the quality of the transmission is highly dependent on frame loss, latency, and latency variation (jitter). On the other hand, in Web hosting, users are primarily concerned about response time, that is, the time required to download a page. To ensure a consistently high level of service, the provider must monitor and measure these metrics closely. The first chart shows that availability was generally good during the report period. However, the service response was a cause for concern. Service response is represented as a percentage of a user-defined limit. The Response/Limit chart shows that many of the services (service to Denver, NYC, and San Francisco in particular) had response times which exceeded the limit. To further monitor the quality of VoIP, the jitter between L.A. and Denver was measured. The chart shows that jitter was minimal at less than 300 ms.

Since the application responses fail to meet the service level objectives, remedial action must be taken immediately by determining the root cause of the slowdown. This can be done by breaking down the response time into client, server, and network. In other words, where is the bottleneck -- is it the client, the server, or the network? If, say, the network is the major source of delay, then the next step is to identify the segment or the internetwork device (router or switch) which is slow. This is usually done by examining the latency of each hop -- i.e., the link between Denver and Chicago was determined to be the bottleneck.

THE POWER OF TREND ANALYSIS
To provide service level monitoring, a database of the vital historical performance statistics must be kept. This information is also an invaluable planning and marketing tool which may be used to answer questions like:

• What is the growth rate of traffic in the customer's network?
• What are the time-of-day or day-of-week traffic variations?
• How consistent is the traffic volume?
• Which circuits generate the highest volume?
• Which are the overutilized and underutilized circuits?

From the service provider's perspective, the information can be used to optimize the core network. For example, successful service level management can help providers identify bottlenecks and effectively allocate resources based on service trends and utilization patterns -- as well as rapidly detect and correct changes in service quality. From the customer's perspective, resources (e.g., bandwidth) can be deployed where they are most needed. Thus the information originally collected for service level monitoring can now be used by both parties to fine-tune the network to better meet the business requirements of the customer.

An effective service level management solution is an essential tool for today's service provider. By providing accurate performance statistics, you can ensure that the customer's expectations can be met -- gaining critical customer loyalty. In addition, the historical and trend analysis can be used for proactive network management that can benefit both the service provider and the customer. For example, understanding how and when to introduce new managed offerings, identifying the impact of new users and technologies on the delivery platforms, and demonstrating the highest quality of service are just a few of the challenges a service level management solution can help solve.

Fred Engel is senior vice president of engineering for Concord Communications. Concord develops next-generation management software that enables effective e-business. Concord's products ensure fast, dependable performance and 24x7 availability of Web sites and e-business services. For more information, visit the company's Web site at www.concord.com.