Tight analyses of two local load balancing algorithms

被引:34
作者
Ghosh, B [1 ]
Leighton, FT
Maggs, BM
Muthukrishnan, S
Plaxton, CG
Rajaraman, R
Richa, AW
Tarjan, RE
Zuckerman, D
机构
[1] Informix Software, Menlo Park, CA 94025 USA
[2] Yale Univ, Dept Comp Sci, New Haven, CT 06520 USA
[3] MIT, Comp Sci Lab, Cambridge, MA 02139 USA
[4] Carnegie Mellon Univ, Sch Comp Sci, Pittsburgh, PA 15213 USA
[5] AT&T Bell Labs, Murray Hill, NJ 07974 USA
[6] Rutgers State Univ, DIMACS, Piscataway, NJ 08855 USA
[7] Univ Texas, Dept Comp Sci, Austin, TX 78712 USA
[8] Princeton Univ, Dept Comp Sci, Princeton, NJ 08544 USA
关键词
load balancing; distributed network algorithms;
D O I
10.1137/S0097539795292208
中图分类号
TP301 [理论、方法];
学科分类号
081202 ;
摘要
This paper presents an analysis of the following load balancing algorithm. At each step, each node in a network examines the number of tokens at each of its neighbors and sends a token to each neighbor with at least 2d + 1 fewer tokens, where d is the maximum degree of any node in the network. We show that within O(Delta/alpha) steps, the algorithm reduces the maximum difference in tokens between any two nodes to at most O((d(2) log n)/alpha), where Delta is the global imbalance in tokens (i.e., the maximum difference between the number of tokens at any node initially and the average number of tokens), n is the number of nodes in the network, and alpha is the edge expansion of the network. The time bound is tight in the sense that for any graph with edge expansion alpha, and for any value Delta, there exists an initial distribution of tokens with imbalance Delta for which the time to reduce the imbalance to even Delta/2 is at least Omega(Delta/alpha). The bound on the final imbalance is tight in the sense that there exists a class of networks that can be locally balanced everywhere (i.e., the maximum difference in tokens between any two neighbors is at most 2d), while the global imbalance remains Omega((d(2) log n)/alpha). Furthermore, we show that upon reaching a state with a global imbalance of O((d(2) log n)/alpha), the time for this algorithm to locally balance the network can be as large as Omega(n(1/2)). We extend our analysis to a variant of this algorithm for dynamic and asynchronous networks. We also present tight bounds for a randomized algorithm in which each node sends at most one token in each step.
引用
收藏
页码:29 / 64
页数:36
相关论文
共 40 条
  • [1] Afek Y., 1992, Proceedings of the Eleventh Annual ACM Symposium on Principles of Distributed Computing, P35, DOI 10.1145/135419.135430
  • [2] Aiello W., 1993, Proceedings of the Twenty-Fifth Annual ACM Symposium on the Theory of Computing, P632, DOI 10.1145/167088.167250
  • [3] SORTING IN C LOG N PARALLEL STEPS
    AJTAI, M
    KOMLOS, J
    SZEMEREDI, E
    [J]. COMBINATORICA, 1983, 3 (01) : 1 - 19
  • [4] [Anonymous], 1993, P 5 ANN ACM S PAR AL, DOI DOI 10.1145/165231.165252
  • [5] [Anonymous], P FOCS 1989
  • [6] Aspnes J., 1991, P 23 ACM S THEOR COM, P348
  • [7] AUFDERHEIDE FM, 1993, P 20 INT C AUT LANG, P398
  • [8] Awerbuch B., 1994, Proceedings of the Twenty-Sixth Annual ACM Symposium on the Theory of Computing, P487, DOI 10.1145/195058.195238
  • [9] Awerbuch B., 1993, Proceedings. 34th Annual Symposium on Foundations of Computer Science (Cat. No.93CH3368-8), P459, DOI 10.1109/SFCS.1993.366841
  • [10] AWERBUCH B, 1989, P 30 IEEE S FDN COMP, P358