We consider the problem of scheduling tasks with unpredictable service times on distinct processing nodes so as to meet a real-time deadline, given that all communication among nodes entails some (possibly large) overhead. This work is motivated by our effort to build a distributed rendering system that uses a cluster of commodity workstations to improve the performance of multimedia applications that require real-time 3D rendering. In this context, the tasks correspond to sets of scene objects that must be rendered to create a single image while the deadline corresponds to the fixed amount of time e to render each frame. The abstract problem then corresponds to the problem of intra-frame scheduling, that is, how to maximize the likelihood that all rendering tasks will be completed on time for the current frame. In this work, we assume the existence of an inter-frame load-balancing agent; that is, we assume that the overall work is periodically partitioned into a number of tasks with...
Thu D. Nguyen, John Zahorjan