We consider load balancing in the following setting. The on-line algorithm is allowed to use n machines, whereas the optimal off-line algorithm is limited to m machines, for some ffixed m < n. We show that while the greedy algorithm has a competitive ratio which decays linearly in the inverse of n=m, the best on-line algorithm has a ratio which decays exponentially in n=m. Specifically, we give an algorithm with competitive ratio of 1 + 1/2n/m(1−o(1)), and a lower bound of 1 + 1/en/m(1+o(1)) on the competitive ratio of any randomized algorithm. We also consider the preemptive case.We show an on-line algorithm with a competitive ratio of 1 + 1=en/m(1+o(1)). We show that the algorithm is optimal by proving a matching lower bound. We also consider the non-preemptive model with temporary tasks. We prove that for n = m + 1, the greedy algorithm is optimal. (It is not optimal for permanent tasks).
|Title of host publication||Algorithm Theory - SWAT 2000 - 7th Scandinavian Workshop on Algorithm Theory, 2000, Proceedings|
|Editors||Magnús M. Halldórsson|
|Number of pages||11|
|ISBN (Print)||3540676902, 9783540676904|
|State||Published - 2000|
|Event||7th Scandinavian Workshop on Algorithm Theory, SWAT 2000 - Bergen, Norway|
Duration: 5 Jul 2000 → 7 Jul 2000
|Name||Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)|
|Conference||7th Scandinavian Workshop on Algorithm Theory, SWAT 2000|
|Period||5/07/00 → 7/07/00|
Bibliographical notePublisher Copyright:
© Springer-Verlag Berlin Heidelberg 2000.
ASJC Scopus subject areas
- Theoretical Computer Science
- Computer Science (all)