Network Computingwww.networkcomputing.com

But it's not all exciting. Policy management is also about the boring, eyelid-lowering oatmeal sandwich of operational and service-delivery procedures: crossing the T's and blurring the eyes; droning on about daytime versus nighttime gold service; and writing down what the deliverables are, how and where they will be measured, and what configurations represent these services.

Networks are larger and more complex than ever. Part of the added complexity is that users are demanding more from their routers, switches and hubs. They want reliable service, reasonable assurances of security, predictable performance (remember, that's part of what drove the PC revolution) and specialized services, such as videoconferencing, that never fail. All this adds management complexity.

And specialized networks are harder to configure. Because there are more interrelationships, these networks have more knobs and are more difficult to troubleshoot when they do have problems. System configurations -- particularly at the edge -- are morphing rapidly to accommodate new customers and changes in their selected services. To top all this off, add in the scarcity of skilled engineers capable of installing, configuring and managing enterprise-level networks. Houston, we have a problem.

Meanwhile, service providers are looking for ways to squeeze additional revenue streams from their infrastructures. Enterprises want to get more services out of their existing networks, thereby reducing costs. In either environment, more knobs are needed to create these services, and these knobs must be turned more frequently. Historically, such a challenge would be met by extremely skilled individuals writing new scripts or using element management to touch the devices or to generate configuration files for them. This type of "nondynamic" provisioning may work well for stable devices at the core of a network but is more problematic in environments where change is more frequent.

Automated policy-based management is supposed to fix it so networks can be run by fewer -- and less highly skilled -- individuals. It does this by distributing intelligence to managed devices and the management applications, so dynamic-change environments can be operated more easily. Note that some skilled individuals will always be needed to build and operate systems. PBNM systems give individuals who understand networking a longer reach; it does not make their talents obsolete -- we discovered this the hard way while testing PBNM tools.

Bottom line: The promise of policy-based management is that people will be able to deploy more complex services across a wider array of devices with fewer highly skilled individuals. The result is increased revenue and decreased expenses. However, the fly in the peanut butter -- you knew there had to be one -- is that fully realizing this goal is impossible without policy and QoS (Quality of Service) standards, which are still being created (see "No Standards, No Policy, No Management," for a look at the state of standards).