The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. ex. Some numerals are expressed as "XNUMX".
Copyrights notice
The original paper is in English. Non-English content has been machine-translated and may contain typographical errors or mistranslations. Copyrights notice
Les systèmes de mémoire partagée distribuée (DSM) au-dessus du réseau de postes de travail sont particulièrement vulnérables à l'impact du faux partage en raison de leurs frais généraux de transaction de mémoire plus élevés et donc des pénalités de faux partage plus élevées. Dans cet article, nous développons un schéma de gestion de mémoire partagée à granularité dynamique qui élimine les faux partages sans sacrifier la transparence des applications de mémoire partagée conventionnelles. Notre approche utilise un arbre splay threadé (TST) spécial pour la gestion des informations de mémoire partagée et un algorithme de synchronisation de compression de chemin basé sur des jetons dynamiques pour le transfert de données. La combinaison du TST et de la compression de chemin est assez efficace ; asymptotiquement, dans un système à n processeurs avec m segments de mémoire partagée, la synchronisation d'au plus s segments prend O(s enregistrer m enregistrer n) amortit les étapes de calcul et génère O(s enregistrer n) messages de communication, respectivement. Sur la base du schéma proposé, nous avons construit un prototype expérimental DSM composé de plusieurs ordinateurs Pentium connectés par Ethernet et exécutant Linux. Les résultats préliminaires de référence sur notre prototype indiquent que notre programme est assez efficace, surpassant largement les programmes traditionnels et pouvant être étendu à grande échelle.
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
Copier
Alexander I-Chi LAI, Chin-Laung LEI, Hann-Huei CHIOU, "A False-Sharing Free Distributed Shared Memory Management Scheme" in IEICE TRANSACTIONS on Information,
vol. E83-D, no. 4, pp. 777-788, April 2000, doi: .
Abstract: 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-memory applications. Our approach utilizes a special threaded splay tree (TST) for shared memory information management, and a dynamic token-based path-compression synchronization algorithm for data transferring. The combination of the TST and path compression is quite efficient; asymptotically, in an n-processor system with m shared memory segments, synchronizing at most s segments takes O(s log m log n) amortized computation steps and generates O(s log n) communication messages, respectively. Based on the proposed scheme we constructed an experimental DSM prototype which consists of several Ethernet-connected Pentium-based computers running Linux. Preliminary benchmark results on our prototype indicate that our scheme is quite efficient, significantly outperforming traditional schemes and scaling up well.
URL: https://global.ieice.org/en_transactions/information/10.1587/e83-d_4_777/_p
Copier
@ARTICLE{e83-d_4_777,
author={Alexander I-Chi LAI, Chin-Laung LEI, Hann-Huei CHIOU, },
journal={IEICE TRANSACTIONS on Information},
title={A False-Sharing Free Distributed Shared Memory Management Scheme},
year={2000},
volume={E83-D},
number={4},
pages={777-788},
abstract={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-memory applications. Our approach utilizes a special threaded splay tree (TST) for shared memory information management, and a dynamic token-based path-compression synchronization algorithm for data transferring. The combination of the TST and path compression is quite efficient; asymptotically, in an n-processor system with m shared memory segments, synchronizing at most s segments takes O(s log m log n) amortized computation steps and generates O(s log n) communication messages, respectively. Based on the proposed scheme we constructed an experimental DSM prototype which consists of several Ethernet-connected Pentium-based computers running Linux. Preliminary benchmark results on our prototype indicate that our scheme is quite efficient, significantly outperforming traditional schemes and scaling up well.},
keywords={},
doi={},
ISSN={},
month={April},}
Copier
TY - JOUR
TI - A False-Sharing Free Distributed Shared Memory Management Scheme
T2 - IEICE TRANSACTIONS on Information
SP - 777
EP - 788
AU - Alexander I-Chi LAI
AU - Chin-Laung LEI
AU - Hann-Huei CHIOU
PY - 2000
DO -
JO - IEICE TRANSACTIONS on Information
SN -
VL - E83-D
IS - 4
JA - IEICE TRANSACTIONS on Information
Y1 - April 2000
AB - 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-memory applications. Our approach utilizes a special threaded splay tree (TST) for shared memory information management, and a dynamic token-based path-compression synchronization algorithm for data transferring. The combination of the TST and path compression is quite efficient; asymptotically, in an n-processor system with m shared memory segments, synchronizing at most s segments takes O(s log m log n) amortized computation steps and generates O(s log n) communication messages, respectively. Based on the proposed scheme we constructed an experimental DSM prototype which consists of several Ethernet-connected Pentium-based computers running Linux. Preliminary benchmark results on our prototype indicate that our scheme is quite efficient, significantly outperforming traditional schemes and scaling up well.
ER -