This paper studies how to exploit the predicted information to maximizeenergy efficiency (EE) of a system supporting hybrid services. To obtain an EEupper bound of predictive resource allocation, we jointly optimize resourceallocation for video on-demand (VoD) and real-time (RT) services to maximize EEby exploiting perfect future large-scale channel gains. We find that theEE-optimal predictive resource allocation is a two-timescale policy, whichmakes a resources usage plan at the beginning of prediction window andallocates resources in each time slot. Analysis shows that if there is only VoDservice, predicting large-scale channel gains and distribution of small-scalechannel gains are necessary to achieve the EE upper bound. If there is only RTservice, future large-scale channel gains cannot help improve EE. However, ifthere are both VoD and RT services, predicting large-scale channel gains ofboth kinds of users are helpful. A low-complexity is proposed, which is robustto prediction errors. Simulation results show that the optimal policy issuperior to the relevant counterparts, and the heuristic policy can achievehigher EE than the optimal policy when the large-scale channel gains areinaccurate.
展开▼
