论文部分内容阅读
本文介绍了一种基于二级网络互联、二级共享存储、三级Ｉ／Ｏ通道的可缩放共享存储多机系统ＳＭＭＰ（ＳｃａｌａｂｌｅＭｅｍｏｒｙ－ｓｈａｒｅｄＭｕｌｔｉ－Ｐｒｏｃｅｓｓｉｎｇ）体系结构。另外还详细介绍了具有信息就地加工能力的共享存储器的设计。由于共享存储器不仅具有存储信息的能力，而且能对有关信息进行就地处理，所以利用该特点能方便有效地实现操作系统中的同步与互斥原语，同时也为某些应用工程中的并行、迭加运算提供了极其有效的操作。经初步测算表明该系统并行与同步效率高，Ｉ／Ｏ吞吐量大，具有较高的性能／价格比，可应用于科学计算、图象处理、三维实体造型、地震数据处理等领域，亦可用作客户机／服务器网络环境下的超级网络服务器。
This paper introduces a scalable shared memory multi-machine system SMMP (Scalable Memory-shared Multi-Processing) architecture based on secondary network interconnection, secondary shared storage, and tertiary I / O channels. In addition, the design of shared memory with information processing capability is introduced in detail. Because shared memory not only has the ability to store information, but also to deal with the relevant information in situ, so the use of this feature can be easily and effectively achieve the operating system synchronization and mutually exclusive primitives, but also for some applications parallel Overlay operations provide extremely efficient operation. The preliminary calculation shows that the system has high parallelism and synchronization efficiency, large I / O throughput and high performance / price ratio, and can be applied in the fields of scientific computing, image processing, 3D solid modeling, seismic data processing and the like. Used as a super web server in a client / server network environment.