VSAM Data Set Design Parameters

(V S A M) is followed b y a qualitative discussion of performance expectations. V S A M data-set design parameters are discussed with respect to performance tradeofs. Analytic techniques are developed for relating some of the VSAM performance sensitivities to data set design parameters. The Virtual Storage Access Method (VSAM)' has been developed for use with virtual storage operating systems. VSAM grew out of the need for an access method that allows data to be accessed both directly by key and sequentially in key-defined collating order. Conventional index-sequential access methods that satisfy this need usually use a chaining technique to insert additions into a file after it has been initially loaded. With these techniques, performance degrades rather substantially as more and more additions are made. VSAM has been designed to avoid performance degradation while retaining the index-sequential facility. Two new logical concepts defined in VSAM are used to manage the space associated with data: the Control Area (CAI; and the Control Interval (crivv). An index is used to address the records contained in control areas and control intervals. An insertion technique is used that works well even after the file has had many records added. The result is a sequential direct insertion facility that-compared with conventional chaining techniques performs well and continues to do so as the fiie is built up. Although VSAM has been designed for use with virtual storage operating systems, it may also be used with all of the os/370 operating systems. Our purpose Iiere is to provide concepts to consider when designing a VSAM data set. This paper describes VSAM and then uses that description to make some qualitative statements about VSAM perforniance expectations as compared with other methods. Some performance tradeoffs are discussed with respect to VSAM data set design parameters. Finally, analytic techniques for some of the crucial VSAM performance effects are devel