When the gain of predictive resource allocation for content delivery is large?

By predicting future information such as data rate with sensory wireless data, radio resources can be pre-allocated for content delivery. Such an integrated sensing and communications technique can help improve network performance and user experience. To justify the cost paid for predicting future information, it is important to understand in which scenarios predictive resource allocation yields a large gain over the non-predictive counterpart. In this paper, we strive to identify the key factors that affect the gain of predictive resource allocation by deriving the closed-form expression of the gain. We are concerned with minimizing the transmission time required for content delivery such as file downloading to users with an expected deadline, where the resources of base stations are shared with real-time services. Then, the performance gain is measured by the difference of the average time required by predictive and non-predictive resource allocation. Inspired by the solution of the optimization problem, we resort to the theory of order statistics for deriving the performance gain. We find that the gain depends on the statistics (i.e., mean value and standard deviation) of the user’s average data rates in the prediction window. Then, we separately analyze how the statistics of the bandwidth available for content delivery and user mobility affect the gain. We use simulation with a real dataset of traffic load to validate the analysis and quantify the impact of the key factors. Our results show that predictive resource allocation can reduce the transmission time even for non-moving users. The performance gain is high when the network is busy or the cell radius is large.