Novel Intelligent Real time Embedded Computing Scheduler for High System QoS quality of service

Abstract

In quest for quality, the researchers are finding ways to maximize the quality of service of a soft real-time embedded computing systems like routers. In this research work, we propose, develop and investigate a rigorous and probabilistic framework of a novel, machine intelligent Quality-of-service Enhanced Stochastic (QUEST) optimal embedded computing packet scheduler for IP and LTE routers. The aim is to enhance the QoS of the system considering the constraint that the processor utilization is reserved at a value which is very close to 100 percent. In this work, maximizing and maintaining the utilization set at a value of nearly 100 percent, we dynamically enhance the system QoS. The unique attributes of the QUEST scheduler are three-fold. First, the QUEST overcomes the challenging issue of the starvation for which the low priority processes suffer. Second, it addresses the challenging issue, which is the failure of the of the standard Earliest Deadline First (EDF) scheduler running at heavy load condition. Finally, the proposed scheduler has the advantage of the pre-programming (arbitrarily) the ratio of the process utilization. In this service-differentiated scheduling framework, we accomplish and maintain a practical case of target steady state process utilization ratio (Pareto) in the ratio of [0.8:0.16:0.04] for the three class of traffic. newlineThe proposed embedded computing scheduler has the characteristics of the re-configurability and it is adaptive in nature. In this scheduling mechanism, we apply a machine-learning based feedback controller for the implementation of the said re-configurability and adaptability. While running, this feedback-controller with the support of the deadline-miss and the cache-miss error feedbacks performs the process of learning and applies the required remedial decisions in order to enhance the QoS of the system. newlineWe investigate important QoS metrics: packet loss rate (PLR), mean waiting time, and scheduling jitter. For performance analysis of the proposed QUEST scheduler, we tak