Near-optimal policies for broadcasting files with unequal sizes∗

Information broadcasting is an effective method to deliver popular files to a large number of users in wireless and satellite networks. In a previous work, we used a dynamic optimization approach to address the problem of broadcast scheduling for a pull system with equal file sizes. In this paper, we address that problem in a more general setting where the file sizes are not equal and have geometric size distributions with possibly different means. The dynamic optimization approach allows us to find a near-optimal scheduling policy, which we use as a benchmark to evaluate a number of other heuristic policies. Also, we modify the resulting policy and apply it to the case with fixed (unequal) file sizes and compare the results with some other well known, as well as new, heuristic policies. Finally, we introduce a low-complexity heuristic policy to be used for practical implementations. The results show that the performance of the new policy is very close to that of the original policy.