Section: New Results
Middleware systems for computational grids
Application deployment on computational Grids
Participants : Boris Daix, Christine Morin.
Computing is an important lever in scientific, technical, and industrial development. However using high performance systems remains a complex activity. Indeed, allocating some computing resources for a given application, installing it on the allocated resources, and executing it is a difficult process for most users of these systems. Automating this process, called deployment, allows to better separate the preoccupations of the system users, of the application developers, and of the infrastructure administrators. This automation also allows to aim at better performance for the applications. Based on a multi-level abstraction model, ODD/SAMURAAIE (On Demand Deployment/System Abstraction Model for User, Resources, And Applications (Actions on, Instance of, and Events from)), our contribution is an automatic deployment architecture [13] . In 2009, we have finalized the Ange proof of concept implementation this architecture. Three significant deployment cases have been automatized thanks to the ODD/SAMURAAIE automatic deployment architecture. This architecture is based on a multi-level system abstraction model (SAMURAAIE) and provides a set of concurrent actors to manage on-demand deployment (ODD) for high-performance computing applications. The deployment cases that have been studied are a parallel application on a remote computer, a master-slave application on a remote computer cluster, and a workflow application on a remote computer grid. In the first case, the application is featured as a directory of compiled files and the computer is multi-processor. In the second case, the application is featured as a compiled file archive and the number of available computers in the cluster changes dynamically. In the third case, the workflow application is featured as a source code archive, contains three sequential codes and a parallel code, and the grid contains two heterogeneous computer clusters. With their growing complexity, the automation of these cases has shown how pertinent are the system abstraction and the on-demand deployment models of the ODD/SAMURAAIE architecture.
A SALOME numerical simulation has been deployed thanks to ANGE and the ODD/SAMURAAIE automatic deployment architecture. SALOME is a component-based, supervised numerical simulation platform. It currently requires that its user allocates some computers, installs SALOME components on them, and declares these "SALOME resources" to the platform itself, all that by hand. After having analyzed this deployment approach, ODD translators and connectors have been developed so that Ange can automatically deploy SALOME simulations on remote infrastructures. A preliminary experiment has shown how SALOME can delegate, to Ange with SALOME support, the deployment of a simple numerical simulation. Future work will improve the internal deployment approach of SALOME so that it can benefit of all automatic deployment features that Ange offers.
This work is carried out in collaboration with Christian Perez from GRAAL project-team. Since April 2009, this work is pursued at EDF R&D where Boris Daix has been recruited.
Adaptation for data management
Participants : Françoise André, Mohamed Zouari.
The usage of context-aware data management in mobile environments has been investigated by Françoise André in collaboration with Mayté Segarra and Jean-Marie Gilliot from Telecom Bretagne Brest (previously known as ENST Bretagne). A context-aware data replication and consistency system that adapts dynamically to changes in the environment has been proposed, based on the use of the Dynaco framework. This work has been supported by a contract (ReCoDEM ) between ENST Bretagne and Orange Labs (previously known as France-TélécomR&D).
In the ReCoDEM project, the distributed aspects of the adaptation system have not been thoroughly investigated. Therefore, a new subject is launched since October 2007 (with M. Zouari as PhD student) to propose a generic distributed adaptation framework. This work focuses on data management in grid and mobile environments; an ambient assisted living application illustrates the approach. Two architectural models have been proposed, one for distributed data management and the other for distributed dynamic adaptation. The models have been implemented as two component-based frameworks which are currently under evaluation. This work is realized in the context of the ALORAD project (Architecture LOgicielle pour la Réplication Adaptative de Données), financed by the Brittany council. Mayté Segarra from Telecom Bretagne Brest is co-adviser for the PhD thesis of M. Zouari.
Adaptation for Service-Oriented Architectures
Participants : Françoise André, Guillaume Gauvrit, Erwan Daubert, André Lage, Nikos Parlavantzas, Jean-Louis Pazat, Chen Wang.
Service-Oriented Computing is a paradigm that is rapidly spreading in all application domains and all environments - grids, clusters of computers, mobile and pervasive platforms. The necessity of dynamic self-adaptation of services to satisfy the different users needs on constantly changing environments is evident.
The following works take place in the context of the S-Cube European Network of Excellence
Services adaptation in distributed and heterogeneous systems
Based on our experience on components adaptation, Françoise André, Guillame Gauvrit and Erwan Daubert are now studying the problems of services adaptation in distributed and heterogeneous systems. This work takes place in the context of the S-Cube European Network of Excellence and covers different aspects such as structural, behavioral and environmental adaptation, distributed decision and planification of adaptation actions, adaptive allocation of resources for services.
Local Service Adaptation
Jean-Louis Pazat and André Lage are studying how to apply dynamic adaptation principles at the level of one single service running on a Grid infrastructure. We are building a prototype based on the OSGi component framework, the XtreemOS (XO SAGA) interface and the Wildcat monitoring tool.
We intend to define explicit links between QoS of OSGi services and Grid resource utilization in order to be able to optimize resource usage while conforming to some SLA.
Dynamic Adaptation of Chemical services
Jean-Louis Pazat and Chen Wang are studying how to express and implement dynamic adaptation of services within the context of chemical programming. We are also working on extensions of the HoCl compiler in order to remove some limitations and add new dynamic features.
