Research On Context-aware HPC Service Discovery Technology

High Performance Computing Services is an important trend of the development of High-Performance Computing. With the number of available high-performance computing servicesincreasing, how to find suitable services become an important issues to be solved. Because of thecharacteristics of high-performance computing services and the diversity of user needs, existedservice discovery methods didn`t meet user discovery goal for high performance computingservices. In this paper, we introduced Context-Aware technology into the field ofhigh-performance computing services and provided users with accurate, personalized discoveryof high-performance services.Context Modeling is the basis for Context-Aware. Based on the analysis ofHigh-Performance Computing services, we constructed the ontology of OWL-HPContext,including a top-level ontology and a middle layer ontology, which provided a unified semanticdescription and classification for context information. The use of context rules to achieve theperception of demand for services is an important method for context-aware, how to obtaincontext rules is the most important issues that need to be addressed. Outside of the manualformulation of rules, a rough set based automatic rule generation method is provided, and inaccordance with the different rule generation types identified rules. A perception of demand forservices method is provided, which including semantic rule matching method and rules conflictresolution method, and it provided accurate, personalized service request information. Accurateand personalized service discovery, on the one hand, depends on service request information, onthe other hand depends on the matchmaking and sorting of services, A suitable service discoverymethod is provided, which included service function matchmaking and personalized servicesorting. Finally, a prototype system is designed and implemented, and the experiments resultsshow that Context-Aware high-performance service discovery method can achieve personalizedsort of services and meet the needs of function requirement at the same time.