Multi-Job Class Queueing Networks With Blocking

在计算机系统、生产系统、通信网络和柔性制造系统等系统的设计和实现中，性能是一个主要问题。这类系统的成败取决于人们对绩效衡量的关注程度。在过去的二十年里，已经多次证明，通过排队网络模型可以很好地评估或预测性能。对于具有无限容量节点的排队网络，在过去的二十年里已经引入了许多方法。然而，由于实际系统中节点的容量是有限的，因此需要使用具有阻塞的排队网络来进行准确的建模以进行性能分析。因此，阻塞排队网络成为近年来一个有趣的研究课题。几位研究人员提出了不同的解决方法。然而，以前的研究仅限于指数服务时间分布、单作业类别和FCFS调度规则。在实际系统中，设备可能具有不同的调度规则，作业行为可能不同，并且服务时间以非指数方式绘制。因此，本项目研究具有有限容量、四种不同节点类型和多个作业类别的排队网络。研究了求解这些模型的几种近似方法和精确解。

Performance is a major issue in the design and implementation of systems such as computer systems, production systems, communication networks, and flexible manufacturing systems. The success or failure of such systems is judged by the degree to which performance measures are of great interest. In the last two decades it has been demonstrated several times that performance can be evaluated or predicted well by queueing network models. For queueing networks with nodes of infinite capacity numerous methods have been introduced in the last two decades. However, since in actual systems nodes have a finite capacity, queueing networks with blocking should be used for accurate modeling for performance analysis. Queueing networks with blocking have thus become an interesting research topic in recent years. Several researchers have suggested different solution techniques. Previous studies, however, are restricted to exponential service time distributions, single job classes and FCFS scheduling disciplines. In actual systems the devices may have different scheduling disciplines, the job behavior may vary and the service times are drawn non-exponentially. Therefore, this project investigates queueing networks having finite capacity, four different node types and multiple job classes. Several approximate ways as well as exact solutions to solving these models are studied.