Reallocation Strategies for User Processing Tasks in Future Cloud-RAN Architectures

In this paper we evaluate strategies to reduce the required processing capacity in a Cloud-Radio Access Network (C-RAN) architecture by improving the placement of user processing tasks. Our approach of assigning compute tasks in a pool of compute resources is based on fine granular tasks, where one compute task per served user is introduced. We compare different strategies in order to balance the load in the pool and save processing resources. Therefore we evaluate the best possible reallocation method by formulating an optimization problem including extensions to reduce the number of reassignments. We also introduce an algorithm for dynamic reallocations that can be implemented in real systems. From the evaluation results we can conclude that all strategies reduce the total overload by enhanced load balancing. Further all strategies improve the perceived Quality of Experience (QoE) of individual users.