Client-side caching strategies and on-demand broadcast algorithms for real-time information dispatch systems

In this work, we propose a broadcast algorithm called Most Request Served (MRS) and its variants with caching strategies for on-time delivery of data in Real-Time Information Dispatch System. This family of algorithms consider request deadline, data object size and data popularity in making scheduling decisions. Although previous scheduling algorithms also base on some or all of these attributes to choose the most beneficial data to be broadcast, they did not consider the loss brought by their scheduling decisions. However, MRS considers both gain and loss in making a scheduling decision. We have performed a series of simulation experiments to compare the performance of various algorithms. Simulation results show that our proposed broadcast algorithm not only succeeds in providing good on-time delivery of data but at the same time provides 20% of improvement in response time over traditional scheduling algorithms like First-InFirst-Out (FIFO) and Earliest-Deadline-First (EDF). Simulation results also show that our proposed caching strategy provides further improvement in terms of percentage of requests finished in time over traditional caching strategy like Least Recently Used (LRU). © 2007 IEEE.