Mobile co-ordination: Providing fault tolerance in tuple space based co-ordination languages

In this paper we describe the concept of mobile co-ordination, a general purpose approach to overcoming failure of agents when using distributed tuple spaces. We demonstrate why mobile co-ordination is better than using existing techniques such as transactions, how mobile co-ordination can provide extra functionality in the form of agent wills, and how the framework to provide this can be implemented in Java and can be used with multiple different tuple space co-ordination languages. Experimental results are presented to show performance gains made when mobile co-ordination is used.