Memory coherence in shared virtual memory systems
ACM Transactions on Computer Systems (TOCS)
Implementation and performance of Munin
SOSP '91 Proceedings of the thirteenth ACM symposium on Operating systems principles
An evaluation of software-based release consistent protocols
Journal of Parallel and Distributed Computing - Special issue on distributed shared memory systems
The SPLASH-2 programs: characterization and methodological considerations
ISCA '95 Proceedings of the 22nd annual international symposium on Computer architecture
Tradeoffs between false sharing and aggregation in software distributed shared memory
PPOPP '97 Proceedings of the sixth ACM SIGPLAN symposium on Principles and practice of parallel programming
Memory consistency and event ordering in scalable shared-memory multiprocessors
ISCA '90 Proceedings of the 17th annual international symposium on Computer Architecture
Software DSM Protocols that Adapt between Single Writer and Multiple Writer
HPCA '97 Proceedings of the 3rd IEEE Symposium on High-Performance Computer Architecture
Efficiently Adapting to Sharing Patterns in Software DSMs
HPCA '98 Proceedings of the 4th International Symposium on High-Performance Computer Architecture
| Hi-index | 0.00 |
Recently, many different protocols have been proposed for software Distributed Shared Memory (DSM) that can provide a shared-memory programming model for distributed memory hardware. The adaptive protocols of these protocols attempt to allow the system to choose between different protocols based on the access patterns it observes in an application. This paper describes several problems that deteriorate the performance of a hybrid protocol[6], an adaptive invalidate/update protocol. To address these problems, this paper then presents a working-set based adaptive invalidate/update protocol that uses a working-set model as the criteria for determining whether to update or invalidate. The proposed protocol was implemented in CVM [7], a software DSM system, and evaluated using eight nodes of an IBM SP2. After experimenting with various working-set window sizes, it was confirmed that the proposed protocol could track an access pattern better than the hybrid protocol, plus with a very small window size the protocol was able to optimize the over-all performance.