An application of hash codes for disk file access, using multiple buckets

Abstract

Hash code methods are widely used for retrieval of information from tables in memory and from direct access storage devices. A key is input to an algorithm which calculates the memory location or the disk address wanted. This paper explains hash code methods for direct disk access by way of a particular application example.The application includes the use of multiple buckets, where each bucket is a separate disk file. Synonym overflow is handled by trying to place a record only once in each bucket (file), and finally placing the record in an overflow bucket (file) if no place could be found in the primary files.The main goal of this design was to utilize 90-95% of the allocated disk space before the average access time became significantly degraded. This is in contrast with the usual requirement for hash code disk access in a single large file that there be about 20% excess space over the amount actually needed.The application was first implemented on an IBM 1130 and was originally conceived to overcome limitations on the size of a single physical file on that machine. It is now running on a Hewlett Packard 3000. The file capacity is 18,000 optometric clinic patient records.