Résumé: Constructing a detailed "analytic" model for a real-world computer system is a complicated task, and interpreting results of apphcatmns may be difficult and risky. This tutorial article presents a case study of a modeling project for a specific IBM 360/65J computer system and describes some practical applications of the resulting central server model. A sketch of the underlying theory of such models is included as an aid to interpreting results. The model was used to obtain the relatmnship of throughput, as measured by CPU activity, to characteristics of the job-stream and to propertms of various hardware and software components of the system. This relationship was used to estimate changes in performance resulting from various alternative modificatmns of the system. The utility of these estimates was confirmed by system measurements taken after some of the recommended moehfications were implemented.

Langue: Anglais
Index décimal 621 .Physique appliquée (électrotechnique, génie civil, génie mécanique, ingénierie appliquée, principes physiques en ingénierie)
Thème Informatique

Mots clés:
Queueing
computer performance prediction
Analytic models
System modifications