Efficient Client Assignment for Client-Server Systems

Many distributed systems use a client-server model in which client assignment strategy plays an important role on the system performance. People use two criteria to evaluate server loads-1) total load and 2) load balance. The total load increases when the load balance decreases, and vice versa. It has been proved that finding the best client assignment is NP-hard. In this paper, we propose a new model for the client assignment problem and design algorithms based on semidefinite programming. We study the identical server case and general server case, present two algorithms (BSP and ABSP), and analyze these algorithms' bounds. In simulation, we evaluate that our client assignement strategies give the satisfiable total load and load balancing using reasonable time compared to the state-of-the-art, thus proving the effectiveness of our algorithms.