A drug discovery grid environment with fault-tolerance support

In order to resolve the long running and error-prone problems in virtual screening, a drug discovery grid (DDGrid) is designed and implemented to establish a grid environment. The environment utilizes idle resources scattered over the CNGrid environment to support high-throughput virtual screening. The DDGrid implements stateful fault detection mechanism based on a complex event processing technology, and provides fault-tolerance support through continuously analyzing the exchanging event streams among different nodes, locating possible faulty situations and triggering the corresponding compensating operations to minimize the possible loss. Experimental results show that the fault detection mechanism keeps the difference between the actual task process time and the theoretical value within 10% through adjusting timeout threshold in the runtime, and improves the usage of underlying computing resources.