An Efficient Deterministic Algorithm for the Resource Discovery Problem
|
Abstract:
We address the problem of how best to get a group of machines on a network to learn of each others existence; this is referred to as the Resource Discovery Problem (RDP). Straightforward algorithms for RDP are slow or have high communication cost. Harchol et al. [3] recently presented name-dropper, a randomized distributed algorithm for RDP which has low time and communication complexity. However, name-dropper has significant limitations--- (1.) the use of randomization precludes it from providing a guaranteed bound on runtime, and, more significantly, (2.) it has no mechanism by which convergence can be detected; in order to provide a high probability of convergence, the number of machines on the network must be known a priori to all machines. We present fast-leader, a deterministic distributed algorithm for RDP which overcomes these limitations while matching name-dropper's time and communication costs.
Citations
| 499 | How to Model an Internetwork – Zegura, Bhattacharjee - 1996 |
| 318 | Consistent hashing and random trees: distributed caching protocols for relieving hot spots on the World Wide Web – Karger, Lehman, et al. - 1997 |
| 251 | An Engineering Approach to Computer Networking – Keshav - 1997 |
| 217 | Frangipani: A scalable distributed file system – Thekkath, Mann, et al. - 1997 |
| 103 | An O(log n) parallel connectivity algorithm – Shiloach, Vishkin - 1982 |
| 63 | Resource Discovery in Distributed Networks – Harchol-Balter, Leighton, et al. - 1999 |
| 59 | E cient Parallel Algorithms for Some Graph Problems – Chin, Lam, et al. - 1982 |
| 53 | A Network Measurement Architecture for Adaptive Applications, Infocom – Stemm, Katz, et al. - 2000 |
