Concurrent data structures

Abstract

Several techniques for the storage of large data structures in main memory have been proposed and, although none is optimal in every situation, tree structures have become a commonly adopted algorithm. A scheme for maintaining a balanced search tree on a distributed memory parallel architecture is described. A general 2P-2-2P (for integer P≥3) search tree is introduced with a linear array of up to [log2N/(P-2)]+1 processors being used to implement such a search tree. As many as ([log2N/(P-2)]+ 1)/2 operations can execute concurrently. Several examples of 2P-2-2P search trees have been implemented on an array of transputer processors. The search structures developed were highly flexible allowing variations in the throughput and response time to be achieved by simple changes to the transputer architecture or the value of P. Applications of these search structures presently in use or under consideration include neural networks, parallel simulation systems, distributed database applications and the migration of sequential systems onto parallel architectures.