Memory Management Issues in Large Scale Shared Memory MIMD Multiprocessors

在紧耦合的MIMD机器中管理各级共享存储的操作系统机制和策略的研究处于早期阶段。对于拥有数百个处理器的大型机器来说，情况更是如此。对这些问题进行了大规模的研究，重点是蝴蝶多处理器。具体项目分为以下几类：跨并行主存储器的共享数据迁移、在每个与主存储器相关联的多个辅助存储器的设置中管理虚拟存储器、文件系统设计以及分布式并发数据结构的设计和评估(这些结构是实施负责共享存储器管理的系统的关键组件)。

The study of operating system mechanisms and policies for the management of various levels of shared storage in tightly-coupled MIMD machines is at an early stage. This is even more true for the case of large-scale machines with hundreds of processors. These issues are investigated in the large-scale with a focus on the Butterfly multiprocessor. Specific projects divide into the following classes: shared data migration across parallel main memories, management of virtual memory in a setting of multiple secondary memories each associated with a main memory, file systems design, and the design and evaluation of distributed, concurrent data structures (which are crucial components in the implementation of systems responsible for shared memory management).