Online Distributed Object Migration

Abstract

We study Distributed Object Migration using competitive analysis. The problem is motivated by distributed object-oriented computing, for which intelligent dynamic migration of (Java or other object-oriented) objects during runtime is important for efficient implementation on multiprocessor systems. In the online version of the problem, k mobile objects reside at n nodes of a network and they respond to a sequence of requests. Each request specifies two objects which have to communicate, and the algorithm has to decide whether to bring the objects together or not. We focus on the case of uniform networks with relatively large communication costs and show tight upper and lower bounds of k, for any network size n≥2. Our algorithm Timestamp uses a timestamp for each object, and we analyze it using an implicit potential function argument; the analysis is interesting in its own right, and may be applicable to a wider class of problems, but it doesn’t seem to be widely used. This implicit potential function argument gives a simple and intuitive proof of the (suboptimal) competitive ratio of 2k−1, within a factor of 2 of the optimal deterministic competitive ratio. To show the optimal competitive ratio of k, we use an explicit, yet less intuitive, potential function.