Skip to Main content Skip to Navigation
Conference papers

Stratégie de découpe de tâche pour le traitement de données massives

Abstract : MapReduce is a design pattern for processing large data sets distributed on a cluster. Its performances are linked to data skews. In order to tackle the latter, we propose an adaptive multi-agent. The agents interact during the job and the dynamic tasks allocation is the outcome of negotiations in order to relieve the most loaded agent and so the running time. In this paper, we show how, when a task is too expensive to be negotiated, an agent can split it in order to negotiate its sub-tasks.
Complete list of metadata

Cited literature [18 references]  Display  Hide  Download

https://hal.inria.fr/hal-01558607
Contributor : Cristal Equipe Smac <>
Submitted on : Saturday, July 8, 2017 - 5:20:45 PM
Last modification on : Friday, December 11, 2020 - 6:44:04 PM
Long-term archiving on: : Wednesday, January 24, 2018 - 6:58:54 AM

File

morge17jfsma.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-01558607, version 1

Citation

Quentin Baert, Anne-Cécile Caron, Maxime Morge, Jean-Christophe Routier. Stratégie de découpe de tâche pour le traitement de données massives. Journées Francophones sur les Systèmes Multi-Agents, Jul 2017, Caen, France. pp.65-75. ⟨hal-01558607⟩

Share

Metrics

Record views

1698

Files downloads

283