Load Balancing White Papers
Parallel Randomized Load Balancing
Overview This paper examined various mechanisms to balance load in a decentralized manner, using the abstraction of a balls and bins occupancy problem. It specifically explored the trade-off between the number of communication rounds invested and the remaining load imbalance. Existing work defined the two extremes in this trade-off, namely, O(log n/ log log n) imbalance after one round, and O(log log n) after n sequential rounds. It demonstrates that, in this models, the sequentiality requirement of the greedy algorithm is in fact necessary for the strong O(log log n) result they obtained.
| Publisher | University of California | File Format | |
|---|---|---|---|
| Date Published | January 2002 | ||
| Format | White Papers | ||
| Topics | |||
-
Did you find this white paper useful? 2 out of 4 users found this white paper useful
White Paper RSS
Download
Register
