Network Computingwww.networkcomputing.com

Yet it is impossible to get a complete look at the services provided to the business user from a single vantage point, drawing on data that's collected far from the end user. Putting management instrumentation out to end nodes is expensive, and may not be justified by the importance of the business process or the revenue it generates.

Simply put, IT professionals need to enable business solutions, not technology. Once the business point of view is identified, objective measures can be weighed (see "Guidelines for Establishing an SLA," at left).

Once the IT department has analyzed the business needs, outside consultants can provide valuable assistance in creating the SLA. But be sure to set realistic expectations: This is the easy stuff compared to monitoring, diagnosing and projecting service levels. Consultants will paint a sweeping picture detailing the process, but they're rarely around when it all hits the fan.

This is not to say that consultants cannot be helpful. In fact, this cross-departmental or interorganizational process might well benefit from a third party's view.

Analyzing the Products Once you know the elements that are most important to your business and how to measure them, and you have decided who will be on your SLA project team, it's time to select your service management tools.

The most comprehensive tools come from the framework vendors, with Hewlett-Packard Co. leading the charge. HP's OpenView division has adopted the mantle of service management through its recent acquisition of Prolin Software and its Service Manager. It is noteworthy for its tight adherence to ITIL (Information Technology Infrastructure Library), a best practice IT services group in the Netherlands. ITIL attempts to cover every possible IT procedure and assign a service value to it. If you're worried about covering every base all the time, this is the approach for you.

However, such frameworks are large, and you'll need help sorting out what's necessary. You'll also need to place a large measure of trust in your management vendor. HP proclaims that it will recommend only those portions of its product line that solve a particular service management need, as defined by a business process. So rather than insist that every solution include OpenView Network Node Manager, for example, HP may suggest only its asset management application.

Another product aimed at defining service management, but one that's less grandiose, is The Empirical Planner from Empirical Software, which offers consulting services with a cookbook-style, do-it-yourself guide for creating SLAs.

End to End Vendors often use the term end to end to describe the totality of an approach. This can be misleading, though, because it implies monitoring, diagnosis and prediction across client, network, server, database and application. But what vendors typically mean is that you'll get end-to-end service management as long as your needs align with the management product's limits.

Compuware Corp.'s EcoSYSTEMS provides a 5,000-foot, real-time, end-to-end view of service management. Compuware expects its products to be deployed by a separate applications monitoring group that serves as a liaison between the network and application development staffs. It does not perform complete system drill-down to find failing hard drives and dirty cache buffers. Rather, it offers a suite of tools for client, network and server approaches to service management. EcoSCOPE, the primary service monitor, watches traffic on the wire, characterizing application traffic by looking at TCP ports, packet headers and open files.

COMMAND/POST from longtime service management vendor Boole & Babbage is a good example of very detailed, complete service management; only the lack of a network component keeps it from being considered completely end-to-end. For many years, COMMAND/POST has monitored Amdahl Corp., Burroughs and IBM Corp. mainframes. As client/server environments entered the picture, COMMAND/POST was extended to reach communications controllers, minicomputers, Unix variants, Novell file servers, Microsoft NT and the desktop. It can even monitor the air conditioners.

Springing up in the shadow of the complex, do-everything world of Boole & Babbage are "glass-to-glass" products such as Pegasus from Ganymede Software, VitalHelp from VitalSigns Software and VeriServ from Response Networks. The glass-to-glass approach refers to response time as measured from the time a query leaves a client's screen until it returns. These products report on the experience from the business user's perspective--truly measuring service where the IT customer cares about it. The downside is that these products focus on monitoring and have very few diagnostic capabilities.