Cognitive Analytics of Social Media Services for Edge Resource Pre-Allocation in Industrial Manufacturing
Dawei Zhu, Zhanyang Xu, Xiaolong Xu, Qingzhan Zhao, Lianyong Qi, Gautam Srivastava · IEEE Transactions on Computational Social Systems · 2021
With the development of industrial intelligence, the resource requests of various social media services in smart cities are expanding rapidly. For hosting services, the edge computing (EC) platform for its low-latency resource provisioning is fully explored. However, the mapping between edge servers (ESs) and services affects the service latency. Meanwhile, the real-time dynamic distribution of resource requirements also impairs the load balance. Therefore, how to optimize the load balance of ESs while meeting the latency-critical requests remains challenging. To deal with the above challenge, in this article, we propose a resource pre-allocation (RPA) method for the social media services with cognitive analytics. Technically, the deep spatiotemporal residual network (ST-ResNet) is employed to complete the cognitive analytics of resource requests. Then based on the analysis results, the optimal resource allocation (ORA) scheme is designed with multiobjective optimization. Finally, the performance of RPA is evaluated by a real-world resource request data set.