Accurate communication cost estimation in static task scheduling

Most of the existing static scheduling schemes either ignore the communication delay or use a very simple model to estimate it. The result is that when the program is actually executed, its execution timing is unpredictable due to network behavior and load. However, if one knows the timing information and communication requirements for the program tasks as well as the network topology and behavior at compile time, one can compute (not estimate) the communication delays and network routes before execution. This information can then be used at execution-time for predictable network timing requirements. The authors show that optimal network routing is NP-complete. They then propose two heuristic algorithms, with different characteristics and complexity, to compute sub-optimal routing information for static scheduling at compile time.<>