Computer network optimization using the power metric for multiple flows: Part I

Abstract

With the rapid expansion of networks and increasing traffic, optimizing network performance has become increasingly important, especially in balancing two competing objectives: increasing throughput and decreasing delay. This paper adopts the Power metric to address this tradeoff, extending the analysis to a general multi-flow model and examining the influence of different queueing disciplines. We introduce three forms of power metrics–individual power, sum of powers, and average power–to capture performance in a multi-flow context. Individual power optimizes each flow’s end-to-end performance, while sum of powers and average power provide a system-wide perspective. These three power metrics are analyzed and optimized under an M/M/1 queueing systems setting, considering two extreme flow discrimination priority disciplines–First-Come, First-Served (FCFS) and Head-of-Line (HOL)–to capture their discriminatory effect on response time while maintaining power optimization. This work is a first step in examining the tradeoff of throughput and delay in queueing systems from various perspectives and across different priority group disciplines. The optimization results aim to provide theoretical insights and guidance for system designers in performance optimization.