Efficient distributed Synchronization within an all-software DSM system for clustered PCs1
نویسندگان
چکیده
We propose a new extension to Chang, Singhal and Liu's scalable algorithm for distributed mutual exclusion in order to allow more than one process at a given time to enter a distributed critical section. This is done by integrating distributed single writer / multiple reader semantics into the original algorithm without affecting the performance for exclusive lock requests. We show that software subsystems providing the shared virtual memory abstraction as well as parallel applications built on it benefit from the new extension significantly. First implementation results employing a newly developed all-software, page based distributed shared memory (DSM) system for clustered Windows NT systems are presented.
منابع مشابه
Efficient Coherency and Synchronization Management in SCI based DSM systems
The performance of shared memory applications on Distributed Shared Memory (DSM) depends to a great part on the existence of efficient synchronization mechanisms and relaxed consistency models with a small management overhead. Using some of the core features of SCI, remote memory access and atomic transactions, highly efficient solutions for these areas have been developed and implemented withi...
متن کاملHardware Supported Synchronization Primitives for Clusters
Parallel architectures with shared memory are well suited to many applications, provided that efficient shared memory access and process synchronization mechanisms are available. When the parallel machine is a cluster with physically distributed memory, software based synchronization mechanisms together with virtual memory infrastructure can implement Software Distributed Shared Memory (S-DSM),...
متن کاملSystem Software Support for Reducing Memory Latency on Distributed Shared Memory Multiprocessors
This paper overviews results from our recent work on building customized system software support for Distributed Shared Memory Multiprocessors. The mechanisms and policies outlined in this paper are connected with a single conceptual thread: they all attempt to reduce the memory latency of parallel programs by optimizing critical system services, while hiding the complex architectural details o...
متن کاملA False-Sharing Free Distributed Shared Memory Management Scheme
Distributed shared memory (DSM) systems on top of network of workstations are especially vulnerable to the impact of false sharing because of their higher memory transaction overheads and thus higher false sharing penalties. In this paper we develop a dynamic-granularity shared memory management scheme that eliminates false sharing without sacrificing the transparency to conventional shared-mem...
متن کاملParallel Pull-Based LRU : a New Request Distribution Algorithm for Large Scale SCI Clustered Web Caches
The SIRAC laboratory has developed SciFS, a Distributed Shared Memory (DSM) tightly integrated with the Linux operating system kernel, that exploits the high performance and remote addressing capabilities of SCI (Scalable Coherent Interface) networks. We experiment with using SciFS to build large scale clustered web caches. An efficient DSM does not only provide a fast and large data storage sp...
متن کامل