International audienceInstitutional desktop grid systems are attractive for running distributed applications with significant computational requirements. While the rapid increasing number of users and applications running on such systems does demonstrate the potential of desktop grid and institutional desktop grid, current implementations follow the old-fashioned master-worker paradigm. Obviously, vulnerability to failures and permanent administrative monitoring are the disadvantages of client-server architectures. To bypass this, we proposed a novel system, called BonjourGrid, able to orchestrate multiple instances of institutional desktop grid middlewares, able to remove the risk of single-source bottleneck and failure, and able to guaranty the continuity of services in a distributed manner. In this paper, we use BonjourGrid protocol, which is based on publish/subscribe paradigm, to show how to adapt it to fulfill all the requirements of a decentralised job scheduler. An evaluation proves that BonjourGrid is able to manage more than 100 applications instanciated in a concurrent way on an institutional desktop grid. Analysing the execution of 100 applications with 2110 tasks during 3 hours demonstrates the potential of BonjourGrid concept and shows that, comparing to a classical desktop grid with one central master, Bonjourgrid gives an acceptable overhead that can be explained
The Desktop Grid technology consists mainly in exploiting personal computer, geographically dispersed, to deliver massive compute power to investigate complex and demanding problems in a variety of different scientific fields. However, as resources number increases, the need for scalability and decentralization becomes more and more essential. Since such properties are exhibited by Peer-to-Peer systems, we aim at using them to create a decentralized desktop grid middleware. Nevertheless, in order to judge the efficiency of such P2P library, an experimental performance evaluation of the provided functionalities is unavoidable. Very few analysis of this kind have been reported, as most evaluations are limited to complexity analysis and to simulations. Such experimental analysis are important, especially when using P2P tools in grid computing context, when applications may have precise efficiency requirement. In this paper, we focus on three libraries: Bonjour, Avahi and Pastry, which provide generic API intended to serve as basis for specialized P2P applications. We perform a performance evaluation of the scalability and their capacity to register and browse an important number of services over 300 hosts in Grid'5000 for recent versions of Pastry, Avahi and Bonjour. We provide detailed analysis explaining the behavior of each library related to two criteria: the elapsed time for registration services and the needed time to discover services. Our aim is to choose the most adequate protocol for creating a decentralized middleware for desktop grid.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.