Performance evaluation of the statistical aggregation by categorization in the SM3 system
Chaitanya K. Baru, S.Y.W. Su · 1984
To perform a statistical aggregation operation over a large file often requires that the records of the file be divided into categories based on the values of the attribute(s) over which some statistical computation is to be performed. It is rather inefficient to perform the necessary data transfer, categorization and statistical computation using a single processor Parallel algorithms designed for multiprocessor systems have been proposed and their performance improvement over the conventional systems has been demonstrated. It is shown in this paper that three to four times performance improvement can be further gained by using a dynamically partitionable multicomputer system with switchable main memory modules (SM3).