Predictive Service Placement in Mobile Edge Computing
Huirong Ma, Zhi Zhou, Xu Chen · 2019
Mobile edge computing is emerging to support delay-sensitive 5G applications at the edge of mobile networks. Unlike traditional centralized clouds, MEC nodes are attached to base stations or access points, thus densely deployed. Therefore, when a user moves erratically among multiple MEC nodes, a new problem arises: whether the services should be dynamically migrated to maintain the service performance (i.e., user-perceived latency). However, frequent service migration can significantly increase operational cost, incurring a conflict between improving performance and reducing cost. To address these mis-aligned objectives, this paper studies the performance optimization of mobile edge service placement under the constraint of long-term cost budget. Aiming at a paradigm shifting from reactive to predictive by leveraging the power of prediction for performance enhancement, we study the predictive service placement with limited prediction of the near-future information. By using two timescale Lyapunov optimization method, we propose a T -slot predictive service placement algorithm (TA) to incorporate the prediction of user mobility based on a frame-based design. We characterize the performance bound of TA in terms of cost-delay tradeoff theoretically. Performance evaluations using realistic data trace show that TA can achieve a superior performance gain over the existing schemes without prediction.