Load Balancing Applying BPN in a Dual-Core Processor in IP Router Running with QUEST Scheduler

Abstract

In this paper, we investigate on a load balancing scheme applying neural network—backpropagation network (BPN) algorithm in a dual-core processor in IP router for processing of multimedia IP traffic. Load balancing is an important challenging issue among the cores of a processor in network routers. A balanced equally distributed and a steady-state load of the cores enhances the system quality of service (QoS) and performance. In this system, each core runs with a QoS-enhanced stochastic smart packet scheduler—QUEST. For analysis, we take both the cases of moderately and heavily unbalanced initial load of the cores of a dual-core processor in router. Each core is provisioned with the QUEST scheduler and runs to attain a ratio of process utilization which is of steady-state featured for scheduling and processing of multimedia IP traffic, namely, deadline-sensitive Voice over Internet Protocol, Internet Protocol Television, and best-effort Hypertext Transfer Protocol. We observe the computed balanced load applying the BPN for different conditions. Applying BPN, the load imbalance among the cores is minimized and the results show that load of each core has attained a target steady-state balanced load.