Massive Mobile Computation Offloading: Operating Data Centers as Virtual Power Plants in Smart Grids

Mobile cloud computing (MCC) is expected to play a vital role due to its capability of scaling the computational resources of mobile devices by offloading tasks to cloud data centers. However, how to jointly optimize the computation offloading in wireless networks and demand response (DR) in smart grids remains open. In this paper, we present a paradigm-shift architecture that operates data centers as virtual power plants, while assuring the Quality-of-Service (QoS) for mobile application. In particular, less tasks are offloaded to a data center when the power gird is heavily loaded, thereby being capable of leveraging batteries of massive mobile devices to compensate the instantaneous power shortage. We formulate a game-theoretic framework for DR-aware dynamic computation offloading that not only guarantees the device lifetime, but also avoids large latency. Mean-field approximation is also applied to tackle the computational complexity when the data center has to schedule a huge number of devices in a distributed manner. Simulation results demonstrate the potential of operating data centers as virtual power plants, where appropriate demand response, balanced device lifetime, and low service latency can be attained simultaneously.