Semantic-based discovery method for high-performance computing resources in cyber-physical systems

Abstract

High-performance computing (HPC) systems with powerful computing capabilities are becoming increasingly significant in the large-scale cyber-physical systems (CPS), helping CPS process a huge number of real-time data. Nowadays, the efficiency of HPC resource management and discovery becomes a challenging problem in CPS, due to the complex characteristics of HPC resources and the growing demands of the users. Although lots of recent efforts have been conducted to the resource discovery in distributed systems, they cannot be well adapted for the cross-regional HPC environments, due to the lack of unified model for the resources description and consideration for the usability demands of non-expert users. In this paper, we propose novel techniques to try to solve the problem. Specifically, we first propose a unified semantic model named HPCRO for specifying cross-regional HPC resources, and apply ontology reasoning to obtain more semantic information for queries. Moreover, we propose a WordNet-based quick resource index list data structure called WQRIL to improve the query. Finally, according to the proposed model and data structure, we propose an efficient discovery method called ROLD for cross-regional HPC resources. Extensive experimental results demonstrate that, our proposals not only maintain efficient resource discovery performance, but also achieve the highest precision rate (94.76%), recall rate (92.34%) and F1-score (93.53%).