Building Quick Resource Index List Using WordNet and High-Performance Computing Resource Ontology towards Efficient Resource Discovery

摘要

High-performance computing (HPC) has played a significant role in scientific discovery and technological innovation, but resources distributed in cross-regional supercomputing centers are still used inefficiently and hard to be managed uniformly, due to various challenging issues like the complex resource types, the heterogeneous interfaces of the resource management software, the inefficient coordination of software and hardware, and the lack of available and standardized description specification model. In this paper, we first propose a HPC resource ontology model called HPCRO to define the concepts of hardware resources and software resources, and specifies types, attributes, logical relations, and data transfer among resources. Moreover, based on the proposed model, we further propose a method called WordNet-based Quick Resource Index List (WQRIL) to support efficient HPC resources discovery. Specifically, the method extends and optimizes the existing method with Quick Service Quick List (QSQL) by extracting synonyms of ontology concepts from WordNet. The method not only can provide a unified representation of HPC resources, but also can support semantic-based fuzzy matching. More importantly, it improves the usability and utilization of hardware and software resources effectively in cross-regional high-performance computing centers. Extensive experiments have been conducted to verify the effectiveness and efficiency of our proposals.