Queueing Dynamics and Optimization Issues in Parallel and Distributed Processing Systems

本课题将从排队理论的角度研究并行和分布式处理系统，并讨论动态优化的相关问题。在基本操作层面出现的并发、协调和调度问题如何影响这些系统的宏观动态和性能的基本理解将被获得。一般分析技术和系统设计原则将被确定。高效操作方案的设计（综合）将被解决，以优化并行和分布式处理系统出现在计算和柔性制造中，关于吞吐量，延迟缓冲占用，铅平衡，容错等标准。从计算和制造的各种实际处理情况中提取的基本原型模型将在统一的框架中进行研究，重点是严格的数学结果和可证明的界限。将开发新的排队模型，以捕捉并发和协调的基本特征，并允许对处理协议的动态和性能进行系统分析。实时调度和资源分配方案将在自然随机框架中进行研究。在相关最优调度算法复杂度较高的情况下，重点将放在探索简单（低复杂度）的调度方案上。其他方面的随机，动态调度也将检查，如果有必要通过模拟。研究应在基本的数学水平上发展对计算机系统和柔性制造中并行和分布式处理的动态方面的理解，并基于坚实的数学方法建立优化性能的技术。该研究的应用分支将研究激励一般模型的具体实际系统。

In this project parallel and distributed processing systems will be studied from a queueing theoretic point of view, and related problems of dynamical optimization will be addressed. A fundamental understanding of how concurrency, coordination and scheduling issues, emerging at the level of basic operations, affect the macroscopic dynamics and performance of such systems will be acquired. General analysis techniques and systematic design principles will be identified. The design (synthesis) of efficient operational schemes will be addressed, for the optimization of parallel and distributed processing systems arising in computing and flexible manufacturing, with respect to criteria concerning throughput, delays buffer occupancies, lead balancing, fault-tolerance etc. Basic prototypical models, extracted from various practical processing situations in computing and manufacturing, will be studied in a unified framework, with emphasis on rigorous mathematical results and provable bounds. New queueing models will be developed to capture essential features of concurrency and coordination and allowing the systematic analysis of the dynamics and performance of processing protocols. Real-time scheduling and resource allocation schemes will be studied in a natural stochastic framework. In cases where the relevant optimal scheduling algorithms are of high complexity, the focus will be on exploring simple (low complexity) scheduling schemes. Other aspects of stochastic, dynamic scheduling will also be examined, if necessary by simulation. The research should develop an understanding of the dynamical aspects of parallel and distributed processing in computer systems and flexible manufacturing, at a fundamental mathematical level, and for establishing techniques to optimize performance, based on a solid mathematical methodology. The applied branch of this research will study specific practical systems that motivate the general models.