High performance distributed systems: resource management and middleware

计算系统和通信网络在合理成本下的可用性正在迅速增加并行和分布式系统的部署。网格提供了对计算、存储、通信和各种其他类型的资源（包括无线设备和传感器）的按需访问，在包括科学、工程和企业计算在内的各个领域受到了极大的关注。实现所需的服务质量水平、可伸缩性、高资源利用率和健壮性对于有效部署此类系统非常重要。有效管理资源对实现这些系统目标至关重要。随着硬件和软件技术的快速发展，异构环境上的互操作性正成为利用前面描述的系统资源的分布式应用程序的重要需求。中间件是提供这种互操作性的粘合剂。由于主要软件公司在Web服务（WS）技术上的数十亿美元投资，未来许多可互操作的系统预计将是基于Web服务的。尽管基于ws的系统可以提供互操作性，但如果设计不当，随之而来的系统开销可能会导致严重的性能损失。提出了网格资源管理以及基于高性能web系统的资源管理研究。研究结果将对网格和其他基于web服务的并行/分布式系统的应用程序设计、操作系统和中间件非常有用。

The availability of computing systems and communication networks at a reasonable cost is rapidly increasing the deployment of parallel and distributed systems. A grid that provides on-demand access to computing, storage, communication and various other types of resources that include wireless devices and sensors is receiving a great deal of attention in various fields including scientific, engineering as well as enterprise computing. Achieving a desired level of quality of service, scalability, high resource utilization and robustness are important for an effective deployment of such systems. Effective management of resources is crucial in achieving these system objectives. With the rapid evolution in both hardware and software technologies inter-operability on a heterogeneous environment is becoming an important requirement for distributed applications that utilize the system resources described earlier. Middleware is the glue that provides such inter-operability. Due to the multi-billion dollar investment in Web Services (WS) technologies by major software companies, many inter-operable systems of the future are expected to be WS-based. Although WS-based systems can provide inter-operability, if not designed carefully, the concomitant system overhead can give rise to severe performance penalties. Research is proposed for resource management on grids as well as on high performance WS-based systems. The research results will be useful in application design, operating systems, and middleware for grids and other web services-based parallel/distributed systems.