Exploring the spectrum of strongly-consistent transactional protocols

Marc Shapiro 1 Masoud Saeida Ardekani 1 Pierre Sutra 2
1 Regal - Large-Scale Distributed Systems and Applications
LIP6 - Laboratoire d'Informatique de Paris 6, Inria Paris-Rocquencourt
Abstract : We explore the spectrum of strong consistency transactional protocols, using both formal properties and performance, in a common framework (based on Deferred-Update Replication) called G-DUR. In G-DUR, a high-performance transactional protocol can be implemented by mixing-and-matching appropriate plug-ins, corresponding to formal properties, in a few hundred lines of code. Therefore, G-DUR enables an apples-to-apples empirical comparison of protocols, and enables studying their limitations and possible extensions, by substituting plug-ins. In this way, we identify four important performance properties. Finally, we describe NMSI, a protocol that satisfies all four properties, and we show that NMSI has the best scalability and performance among strongly-consistent protocols, matching that of a weakly-consistent protocol.
Type de document :
Communication dans un congrès
Workshop on Large-Scale Distributed Systems and Middleware (LADIS), Oct 2014, Cambridge, United Kingdom. Workshop on Large-Scale Distributed Systems and Middleware (LADIS), 2014, Workshop on Large-Scale Distributed Systems and Middleware (LADIS). 〈http://2014.ladisworkshop.org/〉
Liste complète des métadonnées

https://hal.inria.fr/hal-01109740
Contributeur : Marc Shapiro <>
Soumis le : lundi 26 janvier 2015 - 18:56:14
Dernière modification le : vendredi 31 août 2018 - 09:25:54

Identifiants

  • HAL Id : hal-01109740, version 1

Collections

Citation

Marc Shapiro, Masoud Saeida Ardekani, Pierre Sutra. Exploring the spectrum of strongly-consistent transactional protocols. Workshop on Large-Scale Distributed Systems and Middleware (LADIS), Oct 2014, Cambridge, United Kingdom. Workshop on Large-Scale Distributed Systems and Middleware (LADIS), 2014, Workshop on Large-Scale Distributed Systems and Middleware (LADIS). 〈http://2014.ladisworkshop.org/〉. 〈hal-01109740〉

Partager

Métriques

Consultations de la notice

223