A thesis submitter in partial fulfillment of the requirements for the degree of Master of Science, School of Engineering and Technology

Thesis (M.Sc.) - Asian Institute of Technology, 1987

Buffer sizes, the storage and access methods used, cardinality of target tables, and some VM/SP system load indicators were shown to affect the performance of SQL/DS data base and their effects were quantified. Use of appropriate page buffer size and clustering index drastically reduced the CPU time and I/O requirements in retrieval transactions. Regular setting of appropriate page buffer size in order to conserve computer resources and the use the page hit ratio as the parameter to be monitored in order to determine proper page buffer size is recommended. In data base operations involving heavy retrieval transactions from DBSPACEs containing multiple tables, the use of clustering index is re commended to realize savings in computer resources even with the extra DASD space required for the indexes and the inconvenience in creating them. For certain storage/ access cases transaction cost models based on number of pages referenced and cardinality are shown to be in agreement with SQL/DS cost estimate. SQL/DS cost estimate is shown to be a linear function of cardinality in certain storage/access cases. Indexing has minimal effect in reducing transaction cost when the table accessed resides in single-table DBSPACE. Cost reduction is effected when tables have clustering index in multiple-table DBSPACE