USE OF MEAN DISTANCE BETWEEN OVERFLOW RECORDS TO COMPUTE AVERAGE SEARCH LENGTHS IN HAS FILES WITH OPEN ADDRESSING

Abstract

Average search lengths are calculated using both the well known Poisson distribution for the number of addresses assigned a number of records, and an empirical expression for the mean distance between overflow records with the same home address. The method involves computing the number of disk accesses required to randomly retrieve y records overflowing from a given home address, using a knowledge of the mean distance between overflow records. The Poisson distribution is then used to obtain the total accesses required to retrieve all records in the file, from which the average search length may be deduced. The average search length values obtained agree well with the experimental results of Peterson, and the method can be recommended for use in practical design situations. Values of mean distance between overflow records for different loading factors and address capacities are also predicted.