This paper presents a pragmatic approach to providing partition processing system supportfor cooperative applications. A method for specifying and programming application-level partition processing strategies is described. The system support is based on a partition typing mechanism which allows the application programmer to model the relative importance of partitions. This is combined with a split/merge rules configuration table through which the partition processing strategy is defined. In the context of cooperative application semantics, our approach combines tlie correctness of the pessimistic, and the availability of the optimistic approaches for data management in partitionable environments. The paper focuses on the practical issues linked with, firstly, the specification, and secondly, the support at runtime, of the partition processing strategies. This approach is relevant in the context of large-scale asynchronous distributed systems such as the Internet, which, as a result of current technology and topology, are inevitably prone to partitions. Examples are given, illustrating how the partition support is used and combined with new feedback techniques in order to implement more robust cooperative environments.
|Original language||English (US)|
|Title of host publication||Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)|
|Number of pages||18|
|State||Published - Jan 1 1996|