Proactive Management

Because reactive management is challenging at best, administrators are attempting to be more proactiveidentifying indications of potential service disruptions early enough to avert them entirely or to minimize their impact. Proactive management is highly desirable, so many products claim to deliver it. Reduced downtime is one yardstick or metric to measure these claims.

The Benefits of Lead Time

For proactive approaches to yield benefits, they have to provide enough lead time to allow proactive intervention. The lead time must be longer than the time needed to identify the cause and take corrective action; otherwise, the warning appears after the problem has already resulted in service loss. Such a tardy warning may still be useful, as it may speed correct diagnosis, but it would be better if the problem were recognized far enough in advance that it could be avoided entirely. A key here is having tools that shorten the time to produce useful information after a warning is generated.

Proactive analysis tools are therefore driven by continuously monitoring service quality and element behavior. The raw data from the instrumentation system is used to drive a set of real-time functions, such as baselining and predicting future behavior.

Baseline Monitoring

Baselines are continuously calculated based on regular samples of behavioral variables. The baseline is usually calculated with an average or median of the load, a response time, and other monitored attributes. A range defines the normal activity envelope, which is the normal range of the monitored variable through time. The actual behavior is related to the performance envelope; if its value lies within the envelope, the normal range of behavior is being measured.

The trend of the measured value is also important because you want to know if behavior is likely to stay within the expected envelope. If the behavior is trending toward the edge of the envelope, that's more of a concern than if the trend is moving deeper within the envelope.

The baseline is an effective early warning mechanism; however, the warning usually lacks the specific information needed to take specific steps.

The Value of Predicting Behavior

Predicting behavior is essential for making the transition to proactive management. Good predictive tools give administrators the lead time to take steps before a potential problem becomes a service interruption. Predictive tools must be able to do the following:

• Identify potential problems with sufficient accuracy that administrators will heed the warnings and take action

• Provide enough lead time for meaningful responses to be mounted