Performance evaluation of dynamic scheduling algorithms on a multiprocessor cluster

Solutions currently available for task scheduling of multiprocessor systems are mainly static, i.e. they assume a prior complete knowledge of tasks and consequently make a rigid system set-up. These solutions are costly, as they require the design of a specific task allocation scheme for each application, and suffer from inflexibility. The authors simulated performance results of three modified version of the dynamic scheduling bidding algorithm, referred to a multiprocessor cluster. A discrete element dynamic model of the multiprocessor system is applied.