Optimal load balancing algorithms for Jacksonian networks with acknowledgement delays

The load-balancing problem for centralized Jacksonian networks is investigated. The networks are subject to end-to-end flow control with nonzero acknowledgement delay. To analyze these networks, a closed queuing network controlled by a queue with a variable service rate is investigated. The load balancing problem is reduced to one of combined routing and flow control. The objective is to maximize the throughput of the network such that the end-to-end expected time delay does not exceed an upper bound. The optimal routing balances the traffic in the network, and the optimal flow-control policy guarantees the upper bound constraint. Iterative algorithms that derive the optimal routing and flow-control parameters are presented.<>