This paper deals with the application of expert systems to the capacity planning problem, i.e., the process of understanding and predicting the performance of a computer installation to maintain sufficient processing capacity. Such an expert system contains encoded knowledge about computer performance analysis and capacity planning. This is used to interpret measurement and model results and to guide the capacity planning process to a goal consistent with the customer’s data processing requirements.
The Performance Expert Prototype PEP38 is a capacity planning expert system for the IBM System/38. The focus of the system is the effective use of a performance prediction model in the capacity planning process. A blackboard system is used as the communication medium between five modules:
(1) a module that interacts with the user,
(2) the environment module that determines the application workload,
(3) a module that maintains a representation of the computer configuration,
(4) a module that takes a detailed description of the workload and the configuration, and produces various performance predictions, and
(5) an evaluation module that produces an assessment of the current configuration.
A control module directs the execution of the various modules. The authors have found that the expertise required to use a performance prediction model effectively in the configuration design process can be captured and encoded into a knowledge base. Their experience with PEP38 indicates that such an expert system should be able to utilize conventional programs. They found that it is better to use simple rather than complex rules, and the choice of the expert system shell is less important than the overall design structure for the system.
This clearly written paper will be of use to capacity planners. Pertinent references are provided.
