Online Task Admission and Resource Allocation in Edge Computing-Based Cloud Gaming

Cloud gaming refers to an online gaming paradigm that runs game program on remote server and streams the game video back to client device. It allows players to access high-quality games without the need of powerful hardware. The streaming of cloud gaming requires a high-bandwidth, low-latency and stable Internet connection. Conventional cloud gaming is based on cloud computing, but it is gradually being replaced by edge computing. Edge computing is an emerging network computing pattern that deploys edge servers at the edge of the network, close to users. In this way, edge computing is able to provide low-latency and high-bandwidth network computation service, which fits the requirement of cloud gaming perfectly. An edge server needs to serve multiple gaming tasks simultaneously. The CPU and GPU resource requirements of each task varies over time and cannot be predicted accurately. Meanwhile, the arrival and the departure time of a task is unknown until it happens. These online characteristics makes the decision of task admission and resource allocation problem hard to solve. In this paper, the inherited online optimization problem is formulated into a Mixed Integer Nonlinear Programming problem (MINLP). We proposed an online resource reservation and parameter estimation-based greedy (REG) algorithm. According to the result of simulation tests, the performance of the proposed algorithm is 92.12% of the optimal solution.