FlexSched: A Flexible Data Schedule Generator for Multi-Channel Broadcast Systems

This paper is concerned with the issue of building efficient and flexible data broadcast programs for a multiple physical channel environment. The goal of our work has been twofold: (a) to minimize the cost of building the broadcast program and (b) to minimize the average access latency for data requests of mobile clients. To achieve both goals, a highly flexible scheduler, called FlexSched, is proposed, taking a divide-and-conquer approach to solve the data scheduling problem. To determine the optimal allocation of the data to the existing physical channels, FlexSched initially makes a single scan over the vector of access probabilities of the data items and assigns them to the broadcast channels based on the square-root formula of optimal bandwidth usage, i.e., data items are mapped to channels in proportion to the square-root of their access probabilities. For each physical channel a skewed broadcast schedule is then produced by first assigning each data item to its own data group and then iteratively merging neighboring data groups so that the overall average access time of the initial schedule is continously reduced. The merging procedure terminates if there are no more neighboring data groups in the program whose merging would further decrease the overall average access latency. An extensive performance study confirms the effectiveness and efficiency of FlexSched. It is shown that FlexSched significantly outperforms existing data scheduling schemes, while being able to generate the broadcast schedule with an average time complexity of O(N logN).