A comparison of computed chaining to predictors

Computed chaining and predictors are two recent techniques for resolving hashing collisions which use a pseudolink field instead of an actual address link to group records which map to the same home address. With additional computation on the pseudolink, the actual address can be determined. The advantage of the pseudolink is that it often takes much less storage than an actual address would take. This correspondence notes problems with the predictor method which must be overcome if the method is to be used successfully. It also compares the predictor method to computed chaining. The correspondence concludes with a discussion of the utility of multiple predictors, i.e., having more than one chain of pseudolinks for records with the same home address.