Scheduling Multiple Divisible Loads in a Multi-cloud System

In this paper, we propose a novel architecture of a multi-cloud system and investigate the problem of scheduling multiple divisible loads on this system. The scheduling problem takes into account the real-life constraints: the arbitrary release times, i.e., Ready times, of computing nodes, heterogeneous sizes and computation requirement of loads, and the network topology of the system based on dedicated links. We adopt the phase-based multi-round scheduling approach to design two scheduling strategies: a Static Scheduling Strategy (SSS) which assumes that the release times of computing nodes are predetermined and known, and a Dynamic Scheduling Strategy (DSS) which considers that the release times of computing nodes are unknown until they are released. The strategies are designed to achieve high utilization and load balance among computing nodes, thereby minimizing the total processing time of loads. Numerical studies and simulations were carried out to evaluate the performance of the proposed strategies. The results show that the proposed strategies outperform a baseline strategy which does not use the phase-based multi-round scheduling approach.