The Perception and Management of Service Levels

Perception is key

The most critical factor affecting end-user perception of service IT provides is system availability. It's also the factor that most affects end-user productivity, and IT's overall contribution to the corporate bottom line.

Yet sometimes in business, as in life, perception is everything. Let's say that, during the month of May, system data availability and network connectivity were maintained at 100%, and the server was down only 4% of that time. IT management might think that they provided pretty respectable service during May, since two out of the three services were 100% available.

But the end user who sat strumming his fingers on his desk (or 'talking' to his frozen terminal) during those hours probably has a different, and not so favorable, view of IT's service for May.

IT has traditionally measured availability in component availability. But a true measure of service has to be end to end, starting with the end user's experience, through all the technology layers and activities, and back to the end user.

The Details

Measuring availability

The first step in measuring service satisfaction end to end is to define what is a satisfactory service level. As previously mentioned, end users want the applications they need to do their jobs to be available when they need them. If 99% availability of a particular system or set of applications are the negotiated benchmark, be sure that figure's defined within appropriate parameters, like standard hours of operation for that service (even if that's 24x7). And be sure to build into those parameters any planned downtime for maintenance.

After it's clear what service level to measure, the next step is to thoroughly track that service, from when it begins with the end user to when the transaction's completed. What departments/hardware/system infrastructure has a hand in providing the service? Measuring overall service levels means measuring the service that each of these players provides to the overall service level.

Measuring performance

Performance is usually the next most critical factor in making IT customers happy. How end users perceive performance depends on how responsive is the system during interactive transactions (activity between end user and applications) or automatic processing (e.g., running batch jobs). But because both user-initiated and automatic interactions make demands on the same enterprise, the demands overlap. When measuring overall service performance, IT must consider how quickly and efficiently the system processes both interactive and automatic transactions.

IT has a responsibility to balance all the application services provided so no one service uses resources at the expense of other services. But jobs with critical deadlines, such as payroll processing, should take precedence over any batch jobs or even over interactive processing.

Workload levels

How much processing a service performs within a specific time period is a workload level. To get an accurate idea of how well the system is performing and to plan for future capacity requirements, workload levels for the components of the enterprise should be measured. Workload levels can be broken into categories, including:

• transaction rates (both interactive and business transactions)
• number of batch jobs to be processed
• batch jobs completed in parallel

Security

Defining service security in an SLA can be as complicated as the security environment itself. Today's computing environments can comprise applications that cross multiple platforms and can involve users around the world who need access to data across multiple databases. The more complex the system, the more the possibility that providing good system security means using multiple security management systems. And, once again, IT must have a clear understanding of what are the customer's security requirements to know how to best meet them. To accurately define and measure security, a number of components must also be defined, including:

• system users
• system resources
• access control lists, including privilege classes
• intrusion detection

Accuracy

For the purposes of SLM, defining and measuring the accuracy of the data provided by the service is important, but difficult. The following factors directly influence how accurate the data will be:

• data integrity
• data currency
• job control
• scheduled maintenance

To ensure accuracy, these factors must be monitored and checked consistently.

Recoverability

It seems even SLM must follow the laws of physics: what goes up must come down, and every system will occasionally experience an outage. As much as IT hates it, an outage provides an opportunity to shine by providing users with business continuity or as close to continuity as possible. To minimize downtime due to an outage, IT must:

• understand what caused the outage
• understand how the outage will impact end users
• plan for correct recovery procedures depending on the cause of the outage
• recover to the point in time just before the outage
• minimize recovery time

You get what you pay for

It's inevitable that there's tension between the levels of service customers want from IT and the cost of providing that service. Hardware, software and other technologies cost money; the personnel needed to manage that technology cost more money and it is important for users to be aware of this. IT can either have its operating costs be part of the company's general administrative costs or it can charge back the costs of its services to the users of those services.

Charging back for services might make It's customers more aware of the costs of providing the service levels they demand, and might also help customers to view the department as a business partner, rather than a cost center.

When assigning costs to, say, lines of business, considerations include:

• what costs to use
• how to allocate those costs
• how to show customers they're getting value for their money

As the marketplace depends more on electronic commerce, the company IT department is viewed as a competitive advantage. After all, an e-business company with a responsive and highly available Web site will enjoy more sales and greater customer loyalty than a site that is sluggish or unavailable. Since system performance and availability can be directly related to the quality of IT services, charging back for the costs of maintaining that department seems more palatable. And charging for items like memory or software licenses is fairly straightforward. However, charging back for other expenses like utility software is more complex, especially in a distributed environment. It's important that both the IT department and customers agree on what components of service will be charged back to the customer, and fair means of deciding how much to charge.

home