Deep Reinforcement Learning for Online VRC Deployment in Mobile Edge Computing

Deep Reinforcement Learning for Online VRC Deployment in Mobile Edge Computing

연구 분야: Software Development

학회: 2022 IEEE 23rd International Conference on High Performance Switching and Routing (HPSR)

The improvement of communication capabilities and the surge in the number of mobile devices have led to the continuous emergence of a large number of computing-intensive mobile applications. Users have higher requirements for the services provided by the applications, which need high bandwidth and low latency communications. However, the existing traditional network architecture may not be able to achieve this requirement well. As a new network architecture and an extension of cloud computing, Mobile Edge Computing (MEC) is regarded as a promising solution to provide cloud-like computing services for users with low latency. While the number of MEC servers in a region is limited, so for services provider applying virtual machine replicas (VRCs) of the application on the MEC server is a good choice, which give rise to the VRCs deployment problem in the edge network. In this paper, we propose an online deep reinforcement learning method to deploy the VRCs on edge servers so as to minimize the data traffic. Extensive experimental results show that the proposed online deep reinforcement learning method can achieve near-optimal solution with a rather short computing time compared with the enumeration method and have a much better performance from both computing time and transmission delay perspective compared with the available algorithms in the literature.