Performance tuning in OS/VS2 MVS
Thomas Beretvas · IBM Systems Journal · 1978
This paper has presented a practical initial approach to performance tuning; detailed tuning must be tailored to each specific installation. Performance tuning in MVS has become easier through the wider availability of performance tools and the use of an analytical approach to performance evaluation and improvement. Performance tuning has also become much more important because of the cost-consciousness of customers and the heavy resource use of MVS. As a further result of improved analytical capabilities, user expectations are also higher. Thus the conscientious user of MVS has to devote trained analysts to performance monitoring, analysis, and tuning. Such efforts tend to pay for themselves in cost reductions and user satisfaction because of better performance. Although the illustrations provided are mostly for OSlvS2-3.6, the methodology is generally applicable. Tuning involves the minimizing or elimination of unnecessary CPU, 110, and main storage use, the elimination of system bottlenecks, and the possible balancing of the system by favoring one job category at the expense of another, if necessary. Often savings in one resource means additional expenditure in another. Just as often, however, savings in one resource means savings in all. The art and science of performance tuning lie in finding the suitable mix of system adjustments that provides the best available system performance to meet the installation requirements.