article page | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 |

improvement along with a comparative realization of what should be the realistic service goal and therefore how much improvement should be possible. Companies contributing their details to the benchmarking study receive free access to the results. Benchmarking results are available for purchase to other subscribers.

Agents

But management systems - OSS and BSS - have not yet collectively moved towards concrete measures of five-nines. Current best practices and vendor software solutions just do not cut it; new technology and approaches are needed to reliably manage and report on these vast, complex, next generation networks with their rich multi-service platforms. Perhaps the only technological approaches that will punch OSS performance into five-nines Carrier-grade are true service-oriented architectures and autonomic agents. “The general approaches for dealing with flaws are the same for both hardware and software: (1) prediction and estimation, (2) prevention, (3) discovery, (4) repair, and (5) tolerance or exploitation.” [Garland, CMU] Agents offer a convenient level of granularity at which to add redundancy—a key factor in developing robust systems. “Robustness can be achieved through redundancy, and we hypothesize that agents by being naturally smaller and easier to program than conventional systems, are an appropriate unit for adding redundancy.” [CAI03 Autonomic Computing Workshop] Agents, when coupled with service-oriented architectures, designed upon web-services, .NET, or RMI, offer a chance to build and deploy applications and management systems that themselves are assembled like networks are assembled from components. Then, the same principles of design which allow for network components to reach six-nines availability and for networks to perform at full carrier-grade five-nines can be used to assemble networks of intercommunicating software agents.

Last Word

It seems likely that the origin of five-nines as a telecom standard came from analyzing “what can we do?” rather than “what must we do.” Networks, services and equipment were much simpler back then and five-nines while probably a stretch goal, was achievable in equipment. Today’s networks, services, and equipment are much more complicated and contain 1000’s of times more physical components and 100’s to 1000’s of times more software code. Solid state components are individually more reliable and software quality has arguably improved. The expectation for equipment is now six-nines or seven-nines and for software and servers it is

five-nines. For many today these remain stretch goals, but there is no question that everyone in the telecom ecosystem expects and demands high availability. But TL9000 shows us that carrier-grade is more than a measure of availability or any other single quality measure. Carrier-grade is actually an intangible expectation and explicit promise that the equipment vendors will provide the best equipment possible and a clear, immediate communication of issues related to equipment. And that service providers will also provide the best network possible to their customers and keep a clear and immediate communication channel open concerning service impacting situations. And lastly that the supply chain communication is two way, with feedback from the buyer going to the provider so they gauge and support continuous improvement. These behaviors and expectations are captured in the TL9000, particularly its sections on Management Responsibility and Product Realization. So, it becomes important to measure Carrier-grade quality consistently and accurately using the talked about common standards. It also becomes important to measure and provide “true” values for network availability based on service, i.e. the customer experience. The US government has provided a realistic benchmark with its Networx procurement specifications. We suggest that it is a good starting point of well-defined measures and metrics from which to establish a dialogue with customers, vendors and service providers about the fundamental performance requirements, and then aim for perfection. Maybe we will even get that carrier-grade 100,000 year civilization – if we succeed at establishing redundancy of our civilization too.