摘要 近年來網路交換處理系統(switched processing systems)已受到許多學者的關注及討論。環顧幾十年來這領域的研究發展，主要在於如何建構優良的控制策略，使得系統能夠達到最大的吞吐量表現。而最近已備受矚目的MaxProduct這類控制策略，已經能夠使得系統達到最大的吞吐量表現而且保證系統是流量穩定的(rate stable)。但是，當伺服器交換所需時間必須被考量時，經由電腦模擬(computer simulation)的方式可發現，這類控制策略會損失不少的吞吐量。所以本文將提出一演算法，以動態配置佇列長所構成的決策範圍(decision cones)之方式來改善MaxProduct策略的表現。最後，透過電腦模擬的方式我們將展示當伺服器交換所需的時間被考慮時，系統吞吐量的增進成果。 Abstract Switched processing systems have received a lot of attentions in the past years. Over the last decade, research has been focused on how to obtain the maximal throughput of the system by constructing optimal scheduling policies. Recently, a well know class of MaxProduct policies have been shown to maximize the throughput and maintain rate stability of the system. However, simulation studies showed that this policy can lose a huge capacity (throughput) when the switch time between service modes is strictly considered. In this study, we propose an algorithm which modifies the MaxProduct policy by dynamically rearranging the decision cones according to the queue-length states. We also perform a simulation study to show that the proposed approach can diametrically increase the throughput of the system when the switch time is considered.
