Skip to Main content Skip to Navigation
Theses

Support Mémoire Adaptable Pour Serveurs de Données Répartis

Abstract : Data servers are at the heart of information systems (IS) architecture. They consequently must be made of efficient resource management mechanisms and their resource management policies have to be adapted to the behavioral features of the IS.
Unfortunately, it is not possible to build an universal data server that would fit any IS requirements. Actually, ISs differ significantly according to the data they manage, the nature of the applied treatments, guaranteed properties of execution, underlying hardware features, etc. Hence, any IS integrates its own data server with specific resource management policies.
The drawbacks of such an approach are however far from being negligible. First, re-implementing elementary resource management mechanisms is not profitable. Second, because of the rigidity of their architecture, those IS are poorly reactive to (qualitative or quantitative) changes of information, treatments or hardware. Finally, the opaqueness of such systems makes them difficult to coexist on a same platform.
In this thesis, we show that there is no ideal memory management policy. Rather than trying to provide an ideal server, we wish to define a framework that eases the design of adaptable data servers. We focus on the management of the memory of data servers deployed on clusters of workstations. We propose the ADAMS memory support based on an hierarchical management model and event-based communication model. This support eases the integration of different types of memory management policies by distinguishing their respective responsibilities but without doing any hypothesis on their inter-dependencies.
We explain how we integrate ADAMS to the persistent and adaptable distributed shared memory provided by the Arias system. ADAMS extends the features of Arias so as to take into account the specific requirements of data servers and by reducing the granularity of adaptability. We illustrate through an example how the resulting support makes possible the design of data servers that can be dynamically adapted.
Complete list of metadatas

https://tel.archives-ouvertes.fr/tel-00346893
Contributor : Olivier Lobry <>
Submitted on : Friday, December 12, 2008 - 4:12:53 PM
Last modification on : Friday, November 6, 2020 - 4:05:52 AM
Long-term archiving on: : Tuesday, June 8, 2010 - 4:54:35 PM

Identifiers

  • HAL Id : tel-00346893, version 1

Collections

IMAG | CNRS | UJF | UGA

Citation

Olivier Lobry. Support Mémoire Adaptable Pour Serveurs de Données Répartis. Informatique [cs]. Université Joseph-Fourier - Grenoble I, 2000. Français. ⟨tel-00346893⟩

Share

Metrics

Record views

247

Files downloads

97