A linear-time optimal-message distributed algorithm for minimum spanning trees
Distributed Computing
| Hi-index | 0.01 |
Since Gallager, Humblet and Spira first introduced the distributed Minimum Spanning Tree problem, many authors have suggested ways to enhance their basic algorithm to improve its performance. Most of these improved algorithms have even been shown to be very efficient in terms of reducing their worst-case communications and/or time complexity measures. In this paper, however, we take a different approach, basing our comparisons on measurements of the actual running times, numbers of messages sent, etc., when various algorithms are run on large numbers of test networks. We also propose the Distributed Information idea, that yields several new techniques for performance improvement. Simulation results show that contrary to the theoretical analysis, some of the techniques in the literature degrade the performance. Moreover, a simple technique that we propose seems to achieve the best time and message complexity among several algorithms tested.