Ultra lightweight adaptation processes for scheduling servers in push-based systems
Christos Liaskos, Georgios I. Papadimitriou · 2010
Waiting time-optimal scheduling of items for broadcasting in wireless environments requires knowledge of their corresponding access probabilities. Feedback mechanisms have been proposed, according to which the clients notify the server of their needs directly or indirectly. This feedback is then processed at the server, producing the items' access probability distribution function estimate. Based on the principle of maximum uncertainty, a novel, extremely lightweight adaptation scheme algorithm is proposed. Requiring trivial input, this algorithm is shown to estimate almost perfectly the real probability distribution, while alleviating synchronization issues of related approaches and minimizing the processing needs at the server side.