We apply a more realistic cost model than has often been used in the past to the problems of: (1) implementing join algorithms, (2) implementing relational operations which make use of multi-attribute hash files, and (3) creating better multi-attribute hash file organisations using knowledge of the queries which will be asked. This results in a substantial improvement in performance. For some problems a performance improvement of the order of 39 is achieved. We provide preliminary evidence which indicates that even better results are possible.

Equijoin between two relations is one of the basic operations in relational database and a large volume of research have been devoted to it. However, in recent years, there hasn't been a survey which objectively compares a wide spectrum of various join techniques in their relative performances. This survey compares performance and practicality between various join techniques. Main criteria for performance comparisons are disk I/Os. For comparing of practicality, criteria used are easiness and flexibility of implementation. When comparing join techniques, each join technique is evaluated in its full potential, which means that if other techniques are available to enhance this join technique while retaining the main philosophy of it, the later techniques are applied. The main contribution of the paper is that it confirms the believe that no dramatical performance improvement of three major join techniques (nested loops, sort-based, and hash based) of relational database can b...

Abstract not found