Françoise André

Pandore: a system to manage data distribution

The goal of the Pandore system is to allow the execution of parallel algorithms on DMPCs (Distributed Memory Parallel Computers) without having to take into account the low-level characteristics of the target distributed computer to program the algorithm. No explicit process definition and interprocess communications are needed. Parallelization is achieved through logical data organization. The Pandore system provides the user with a mean to specify data partitioning and data distribution over a domain of virtual processors for each parallel step of his algorithm. At compile time, Pandore splits the original program into parallel processes. Each process will execute some appropriate parts of the original code, according to the given data decomposition. In order to achieve a correct utilization of the data structures distributed over the processors, the Pandore system provides an execution scheme based on a communication layer, which is an abstraction of a message-passing architecture. This intermediate level is them implemented using the effective primitives of the real architecture (in our specific case, an Intel iPSC/2).

A framework for dynamic adaptation in wireless environments

The important differences between wireless and wired environments make it indispensable for applications to adapt to the type of environment in which they execute. Solutions provided by existing approaches are often embedded in the heart of the applications functionalities making them more difficult to develop and to maintain. We propose the utilization of an adaptive framework, MoleNE, which provides applications with a set of services as an abstraction of concepts usually utilized to manage mobility issues. Services are built as a set of MoleNE components that can be customized for different applications. Further a MoleNE component can be provided with a special object the Controller, which performs the adaptation of the component behavior depending on the execution conditions. This paper describes the architecture of MoleNE components and gives evidence of the easy use of the adaptation mechanisms by presenting an adaptation strategy for a transactional hospital application.

Dynamic adaptation for grid computing

As Grid architectures provide resources that fluctuate, applications that should be run on such environments must be able to take into account the changes that may occur. This document describes how applications can be built from components that may dynamically adapt themselves. We propose a generic framework to help the developpers of such components. In the case of a component that encapsulates a parallel code, a consistency model for the dynamic adaptation is defined. An implementation of a restricted consistency model allowed us to experiment our ideas.

Supporting adaptable applications in grid resource management systems

Grid computing promises to bring the resources to satisfy the increasing requirements of scientific applications. As grids result from several organizations that pool their computational resources, resource availability varies frequently inside grids. Relying on autonomous dynamic adaptability and managing dynamic collections of resources, technologies have been proposed in order to handle those variations at the level of applications. However, despite applications have evolved in order to fit better dynamic grid environments, grid resource managers still restrict to rigid jobs, thus inhibiting application adaptability and malleability. This paper discusses 3 options to overcome that restriction. Malleable job management can be built on top of existing unmodified infrastructures. It can also be implemented as a modification of the infrastructure. At last, we propose an intermediate approach that fosters the cooperation between the infrastructure and its users. Requiring an initial modification of the infrastructure, the latter design combines cost efficiency with possibility to further extend the job model without any additional modification of the infrastructure. In the discussion, qualitative arguments arc supported by some experimental results.

The Pandore data-parallel compiler and its portable runtime

This paper presents an environment for programming distributed memory computers using High Performance Fortran. Emphasis is put on compilation techniques and distributed array management. Results are shown for some well known numerical algorithms.

Dynamic data replication and consistency in mobile environments

In this article, we investigate the usage of context-aware replication in mobile environments. We present the problems that can occur in such situations and show that existing replication techniques are not well adapted to this problem. For achieving this replication aspect, we propose a context-aware replication system that adapts dynamically the data replication to changes in context informations. We present an internal and external view of this system and the expected adaptation.

Dynamic Adaptation of Parallel Codes: Toward Self-Adaptable Components for the Grid

One of the challenges that come from the emergence of Grid architectures is to invent new programming techniques for these new platforms. As we explain in this chapter, we think that the architecture of the applications should reflect both the parallel and the distributed aspects of Grid architectures. It results in applications built as assemblies of parallel components. Since Grid architectures are known to be highly dynamic, using resources efficiently on such architectures is a challenging problem. Software must be able to react dynamically to the changes of the underlying execution environment. In order to help developers to create software for the Grid, we are investigating a model for the adaptation of parallel components. This chapter focuses on the adaptation mechanisms that are provided as a meta-level for components. We describe how a generic platform can help to develop efficient Grid software. First experimental results show the gain that can be expected from the use of such a platform.

Enforcing consistency during the adaptation of a parallel component

As grid architectures provide execution environments that are distributed, parallel and dynamic, applications require to be not only parallel and distributed, but also able to adapt themselves to their execution environment. This article presents a model for designing self-adaptable parallel components that can be assembled to build applications for grid. This model includes the definition of a consistency criterion for the dynamic adaptation of SPMD components. We propose a solution to implement this criterion. It has been evaluated on both synthetic and real codes to exhibit the behavior of the several proposed strategies

Building a Context-Aware Ambient Assisted Living Application Using a Self-Adaptive Distributed Model

Ambient assisted living (AAL) applications are intended to allow elderly people to stay at home while connected to their family and medical staff. These applications involve several services as health-care services or infotainment, which are executed on different environments such as hospital, proximity doctor, family homes. These environments constitute a distributed, heterogeneous, highly dynamic infrastructure that makes it necessary to build AAL applications as adaptive software including the necessary mechanisms to dynamically modify their behavior. In this paper we propose a multi-level model to build context-aware self-adaptive entities: the basic level defines the mandatory functionalities for adaptive entities, while upper-levels define how to coordinate adaptation of primitive or composite entities. The generic nature of the model is illustrated by its use when building a self-adaptive distributed data management service for an AAL application.

AeDEn : An Adaptive Framework for Dynamic Distribution over Mobile Environments

Mobile computing is a domain in great expansion. Wireless networks (gsm, satellite, etc) and Portable Information Appliances pias (laptops, PDAs, cellular phones, etc) are developing very rapidly. More and more mobile users would like to execute their applications with the same quality of service as on their desktop station, whatever their needs in memory and computing power. Using such applications in a mobile environment raises new challenges. Some of these applications are extremely costly in system and network resources, whereas pias resources are poor and wireless networks offer a very variable quality of connection. In this paper we propose an adaptive and dynamic distribution of applications on the local environment to overcome the poorness of available resources on pias, and to reduce and regulate variability effects. Moreover, due to the variety of distribution policies, we propose a framework providing adaptive distribution policies.RésuméL’informatique nomade est un secteur en pleine évolution aussi bien au niveau des terminaux mobiles (ordinateurs portables, assistants personnels, téléphones portables, etc), qu’au niveau de l’infrastructure et des protocoles des réseaux de communication sans fil (gsm, satellites, etc). De plus en plus d’utilisateurs mobiles veulent pouvoir utiliser leurs applications de la même manière qu’en environnement fixe et cela, quel que soit le volume de données à traiter et quel que soit le temps de calcul demandé. L’utilisation de telles applications à partir de mobiles posent néanmoins certains problèmes. D’un côté, ces applications sont coûteuses en ressources système et/ou réseau, et de l’autre côté, le mobile offre peu de ressources et celles-ci sont fluctuantes. Nous proposons une approche consistant en une distribution adaptative des applications permettant d’utiliser dynamiquement les ressources de l’environnement pour pallier l’insuffisance des ressources du mobile. De plus, face à la diversité des politiques de distribution existantes, nous proposons un cadre général de conception permettant une utilisation de politiques adaptables en fonction de l’environnement.