A Constraint Language Approach to Grid Resource Selection

The need to discover and select entities that match specified requirements arises in many contexts in distributed systems. Meeting this need is complicated by the fact that not only may the potential consumer specify constraints on resources, but the owner of the entity in question may specify constraints on the consumer. This observation has motivated Raman et al. to propose that discovery and selection be implemented as a symmetric matching process, an approach they take in their ClassAds system. We present here a new approach to symmetric matching that achieves significant advances in expressivity relative to the current ClassAds—for example, allowing for multiway matches and querying on policy. The key to our approach is that we reinterpret matching as a constraint satisfaction problem and exploit constraint-solving technologies to implement matching operations. We have prototyped a system, RedLine, based on this approach and have conducted preliminary experiments that suggest that it can be implemented at least as efficiently as ClassAds. We have also successfully applied RedLine to some challenging matching problems from several application domains.