Conngurable Shared Virtual Memory for Parallel Computing
نویسندگان
چکیده
Most distributed-shared memory(DSM) systems use a global address space which is shared by all processing nodes. Although the global address space is convenient in providing a single address space for programming models, it is applicable to DSM systems with relatively small number of processing nodes. As the system size grows, there is a problem with scalability. To avoid this drawback, a new shared virtual memory management scheme, \CSVM (Conngurable Shared Virtual Memory)," is proposed. CSVM creates a new single address space for a given program which is shared only by the nodes cooperating for the parallel execution of the program. The new scheme, therefore, can be applied to any size of DSM system and even to heterogeneous DSM systems. We have developed two algorithms for the distribution and the management of a shared address space. These algorithms cover both the virtual memory of sequential computer and the shared virtual memory of the traditional DSM system, which means CSVM is reduced to the virtual memory management of sequential computer in executing a sequential program and to the traditional shared virtual memory in executing a parallel program. According to the analytical result, CSVM reduces the message traac of DSM system with global address very eeectively.
منابع مشابه
High-Level Parallel Programming of an Adaptive Mesh Application Using the Illinois Concert System
We have used the Illinois Concert C++ system (which supports dynamic, object-based parallelism) to parallelize a exible adaptive mesh reenement code for the Cosmology NSF Grand Challenge. Out goal is to enable programmers of large-scale numerical applications to build complex applications with irregular structure using a high-level interface. The key elements are an aggressive optimizing compil...
متن کاملDistributed Shared Memory
This paper presents the overview of parallel computer models. Practical parallel machines which are parallel virtual machine (PVP), symmetric multiprocessor (SMP), massively parallel processor (MPP), distributed shared memory (DSM), and clusters of workstation (COWs) are described. According to the classification of practical parallel machines, this paper mainly deals with distributed shared me...
متن کاملSemiotics of Collective Memory of the Iran-Iraq War (Holy Defence): A Case Study of the Shared Images in Virtual Social Networks
This study aims to achieve a semiotic understanding of collective memory of the Iran-Iraq war. For this purpose, samples of images in virtual social networks shared in response to the news of discovery and return of the bodies of more than 175 divers have been analyzed. Visual signs in photographs, cartoons, graphic designs, prints, paintings and posters, in methods of historical pictures and f...
متن کاملShared virtual memory clusters: bridging the cost-performance gap between SMPs and hardware DSM systems
Although the shared memory abstraction is gaining ground as a programming abstraction for parallel computing, the main platforms that support it, small-scale symmetric multiprocessors (SMPs) and hardware cache-coherent distributed shared memory systems (DSMs), seem to lie inherently at the extremes of the cost-performance spectrum for parallel systems. In this paper we examine if shared virtual...
متن کاملJavaNow: A Framework for Parallel Computing on Networks of Workstations
Networks of workstations are becoming a dominant force in the distributed computing arena, this due primarily to the excellent price to performance ratio of such systems when compared to traditional massively parallel architectures. It is therefore critical to develop programming languages and environments that can help harness the raw computational power available on these systems. The JavaNow...
متن کامل