Methods and tools for organizing the global job queue in the geographically distributed computing system

The geographically distributed computing infrastructure (DCI) considered in the paper includes high performance computing systems united by communication channels. Computing systems from the DCI are high-performance clusters differing in architecture and performance. Communication channels uniting clusters have different reliability and bandwidth. The considered model of DCI has a decentralized jobs management and dispatching scheme. This scheme implies that at any time malfunction of any computing cluster or a failure in the communication channel can cause cluster’s leaving the DCI. Cluster’s or channel’s troubleshooting means dynamically connecting the cluster to the DCI. The global job queue is organized in this computing infrastructure. Computing jobs have absolute priorities, and high priority job can interrupt low priority running jobs. Jobs from the global queue allocate on idle resources of computing systems. Forming and storing global job queue in conditions of dynamically changing DCI composition needs the reliable information system. The authors reviewed some distributed DBMSs as the basis of this information system. The article outlines the requirements for a distributed information system. The authors conducted a comparative analysis and selected a solution that satisfies the requirements, and designed prototype of the geographically distributed computing infrastructure with the decentralized scheme of jobs dispatching.