Balls into bins via local search: cover time and maximum load


Loading...

Date

2014

Publication Type

Conference Paper

ETH Bibliography

no

Citations

Altmetric

Data

Abstract

We study a natural process for allocating m balls into n bins that are organized as the vertices of an undirected graph G. Balls arrive one at a time. When a ball arrives, it first chooses a vertex u in G uniformly at random. Then the ball performs a local search in G starting from u until it reaches a vertex with local minimum load, where the ball is finally placed on. Then the next ball arrives and this procedure is repeated. For the case m=n, we give an upper bound for the maximum load on graphs with bounded degrees. We also propose the study of the cover time of this process, which is defined as the smallest m so that every bin has at least one ball allocated to it. We establish an upper bound for the cover time on graphs with bounded degrees. Our bounds for the maximum load and the cover time are tight when the graph is vertex transitive or sufficiently homogeneous. We also give upper bounds for the maximum load when m>=n.

Publication status

published

Book title

31st International Symposium on Theoretical Aspects of Computer Science (STACS 2014)

Volume

25

Pages / Article No.

187 - 198

Publisher

Schloss Dagstuhl – Leibniz-Zentrum für Informatik

Event

31st International Symposium on Theoretical Aspects of Computer Science, STACS 2014

Edition / version

Methods

Software

Geographic location

Date collected

Date created

Subject

Balls and Bins; Stochastic Process; Randomized Algorithm; Balls and Bins; Stochastic Process

Organisational unit

03672 - Steger, Angelika / Steger, Angelika check_circle

Notes

Funding

Related publications and datasets