Designing Fault-Tolerant Mobile Systems

The purpose of this paper is to investigate how several innovative techniques, not all initially intended for fault-tolerance, can be applied in providing fault tolerance of complex mobile agent systems. Due to their roaming nature, mobile agents usually run on Java-based platforms, which ensures full portability of mobile code. The rst part of the paper discusses speciic characteristics of mobile systems, outlines the application areas beneeting from code mobility, and shows why the existing error recovery techniques are not suitable for mobile systems. In the next part of the paper we present evaluation criteria for fault tolerance techniques, and propose several possible solutions for error recovery at the application level: meta-agent, Coordinated Atomic actions, asyn-chronous resolution, self-repair, and proof carrying code. The intention is to allow system developers to choose the approach which is suited best to the characteristics of the mobile agent application to be designed. To this end we discuss the advantages and disadvantages of each technique, as well as situations in which it provides the most beneet. A simple example, based on Internet shopping, is used throughout the paper to demonstrate the techniques.