Using a simulation model, experiments were conducted on various directory organization schemes and their performance implications. In particular we tested the effects of a multiprogrammed environment on system throughput for retrieval operations. Analysis of the results shows that different factors are relevant to performance for the various systems, and that under some circumstances ISAM and hash-coding may lose the advantages they possess over B-trees in a stand-alone environment when multiprogramming is used. Some results of sharing data among simultaneous users are also presented.

Performance modelling of data base management systems, and in particular data base models capable of describing representation features in secondary storage, are discussed. Data structures relevant for secondary storage representations are defined, and a partial characterization of several actual systems is given. The accent is on the use of these structures in the context of a simulation model. We sketch the operation of a generalized modelling program in terms of these structures. Data description and system mechanisms required to generate attribute values are treated. The problem of "appropriate values" is discussed in the context of two probabilistic models.

Abstract A design for a time‐sharing monitor on a small‐to‐medium size computer is presented. A system objective is to achieve good resource utilization and thus be able to do more with limited resources. To this end, some assumptions are made about the user environment (information management applications), and some co‐operation is required from application programmers. The design is machine independent, but 360 and Sigma 2 implementations are also described.