A distributed dynamic grid scheduler for mixed tasks

We consider scheduling of bag of independent mixed tasks (Hard, firm and soft) in a distributed dynamic grid environment. Recently, a general distributed scalable grid scheduler (GDS) for independent tasks was proposed to maximize successful schedule percent in an error-free grid environment. However, GDS did not consider constraint failure of task during execution due to resource overload, which leads to limited successful schedule percent. In this paper, we propose a novel distributed dynamic grid scheduler for mixed tasks (DDGS-MT), which takes into consideration the constraint failure of task during execution due to resource overload. The proposed scheduler incorporates migration and resume fault tolerant mechanisms for computational and communication intensive tasks respectively. The proposed scheduler shows improved performance in terms of successful schedule percent and makespan in comparison with GDS. The results of our exhaustive simulations experiments demonstrate the superiority of DDGS over GDS scheduler.