Service level management is again an internal matter

Corporate service-level management is increasingly becoming an inward-looking activity, according to John McConnell, network analyst and president of Boulder, Col.-based McConnell Consulting Inc.

McConnell recently visited Toronto to discuss effective enterprise management through service-level management.

“When we used to think of service providers we would talk about some other group outside the organization…and what we’re starting to see now is a fairly significant shift that internal IT or IS groups are actually being forced to take on this service provider role,” McConnell said. “If you’re not a service provider, you’re going to need to figure out how to play that role because you’re going to be forced into it.”

According to McConnell, the next generation of service providers is emerging, and with this a second look at the function and form of service level management.

“We are starting to see IS organizations again revisiting the issue of how do they start to become revenue generators…and provide all the customers support with competitive advantage.”

Currently, the kind of service level management that organizations are looking for involves the ability to improve communication and increase provider and management team accountability, McConnell said. In addition, people are seeking to verify investments in networking management technology and provide a more effective basis for managing decisions and use of infrastructure.

At the centre of all this lies the service level agreement, what “has sort of been the Holy Grail,” McConnell said. “One of the places people still run aground is the precision of your agreement.”

Among the factors to be considered in any agreement are: the terms being agreed upon, the measurement of service quality, a definition of what constitutes a breach, how the service level will be validated, reporting requirements and penalties.

Research has found that most people are concerned with measuring and managing the level of network availability. While this is necessary, McConnell said it is not sufficient. Moving beyond simple availability means taking a closer look at applications and servers, issues of latency and stability, response time, tools and basic reliability of network connections.

One way to organize these issues is to integrate transport, computing and applications infrastructures, which are “big chunks of the management problem you’ve got to solve anyway,” he said.

When looking at transport infrastructure, it is important to look for vendors that put all information in a single repository while also considering matters of tracing, monitoring, prioritization and resource allocation. Computer infrastructure, which is most important administratively, looks to find clients and servers that support the system and ask the questions: Are they network ready and remotely manageable? Can you apply policies? Can you group the systems?

Most important for the applications infrastructure is how service components are being deployed, the ability to measure behaviour of applications and services, protection and location of data and sychronization and replication of strategies.

Instrumentation is perhaps the most important, because it “is one of the fundamental disciplines. If you can’t measure things you certainly don’t have a prayer of managing them effectively.”

Event management is another. Event collection and filtering can assist with the where and why an event occurs. The real payoff, however, is in event correlation, McConnell said, which is important because “if we’re not careful we waste a lot of time chasing the wrong artefact.”

Other important disciplines include analysis and reporting, which allow fault isolation and case-based reasoning, and “trending,” which finds patterns in events so problems can be viewed as well as forecasted. The key analysis tool, however, is optimization, which assists in deciding if everything is put together as it should be.

Creating policies is also meaningful because not only does it allow control over actions of the management system, but the ability to define actions that support business goals, McConnell said. Policy attributes may include: initiating the user, targeting resources, location, time, precedence, audit or duration.

When all is said and done, building a service level agreement starts with implementation, but also requires introduction and evaluation. Not only is it necessary to phase in agreements for groups, each must be stabilized before moving on to the next. As far as evaluation, McConnell said, one must determine the reality of the agreement and continue to search for existing problems.