QoS Driven Task Offloading and Resource Allocation at Edge Servers in RAN Slicing

Abstract

Task offloading results in the remote execution of tasks, thereby reducing the load on the lower capacity devices and mobile instruments. The offloaded tasks to edge servers in RANs get executed in container-based virtualization technologies. In this paper, we examine traffic offtoading and scheduling, where we investigate QoS-based traffic assignment to edge servers in network slices. We propose an ensemble method for classifying through Multiple Attribute Decision Making (MADM), Single Attribute Categorization (SAC), and fuzzy rules. Then, we apply enhanced weighted Borda scoring to categorize the task into its priority class, which are placed in their respective Kafka topics. Finally, we present a probabilistic, priority-driven Kafka-topic consumer which schedules the offloaded tasks in the edge containers. The slice-based setup constitutes of Flowvisor, Mininet, Beacon and Pox controllers, Kafka, and Docker engine. The proposed ensemble categorization exhibits 26% and 12.5% better accuracy than simple additive and multiplicative exponential weighting MADM methods. Experimental results show that the proposed scheduling methodology on average reduces long piling of medium and low priority tasks by a factor of 7% and 12% respectively.