Future Generation Computer Systems 25 (2009) 747–755 Contents lists available at ScienceDirect Future Generation Computer Systems journal homepage: www.elsevier.com/locate/fgcs An agent architecture for managing data resources in a grid environment María S. Pérez a,* , Alberto Sánchez a , Jemal H. Abawajy b , Víctor Robles a , José M. Peña a a DATSI. FI. Universidad Politécnica de Madrid, Spain b Deakin University, Victoria, Australia article info Article history: Received 20 January 2006 Received in revised form 2 July 2008 Accepted 22 July 2008 Available online 31 July 2008 Keywords: Multiagent system Cooperation Parallel file system Data grid abstract The agent paradigm has been successfully used in a large number of research areas. MAPFS, a parallel file system, constitutes one successful application of agents to the I/O field, providing a multiagent I/O architecture. The use of a multiagent system implies coordination and cooperation among its agents. MAPFS is oriented to clusters of workstations, where agents are applied in order to provide features such as caching or prefetching. The adaptation of MAPFS to a grid environment is named MAPFS-Grid. Agents can help to increase the performance of data-intensive applications running on top of the grid. This paper describes the conceptual agent framework and the communication model used in MAPFS- Grid, which provides the management of data resources in a grid environment. The evaluation of our proposal shows the advantages of using agents in a data grid. © 2008 Elsevier B.V. All rights reserved. 1. Introduction Several agent-based applications have been built in diverse areas such as business [20], electric management [19,7], control [2, 3], networks [11] or in general, industrial applications [17,18]. Agents provide several concepts, which allow software engi- neers to design applications in a way closer to the human thought. Furthermore, agents offer useful characteristics to deal with com- plex and dynamic environments. Nevertheless, if our objective is applying the agent technology to a field, like operating systems, we find some difficulties. In fact, there are important limitations when agents are applied to operating systems. These can make complex the success combination of these two fields. The most important ones are: • The agent paradigm interacts with the infrastructure at a higher level than the operating system. • Efficiency is a very strict requirement in the case of the development of a component of the operating system, and specifically a file system. Agent technology introduces an abstraction layer and, thus, it involves a loss of efficiency. Nevertheless, these limitations can be avoided, since the agent paradigm distinguishes clearly between agent theory, * Corresponding address: DATSI. FI. Universidad Politecnica de Madrid, Facultad de Informatica, Campus de MontegancedoBoadilla del Monte, 28660 Madrid, Spain. Tel.: +34 91 336 73 80; fax: +34 91 336 73 73. E-mail addresses: mperez@fi.upm.es (M.S. Pérez), ascampos@fi.upm.es (A. Sánchez), jemal@deakin.edu.au (J.H. Abawajy), vrobles@fi.upm.es (V. Robles), jmpena@fi.upm.es (J.M. Peña). which provides concepts related to the agent field, and agents architectures, which offer specific solutions and implementations. MAPFS is a successful application of the agent theory in the development of a parallel file system [26]. The same philosophy is applied to the design of MAPFS-Grid [27], the adaptation of this parallel file system to data grids. Our proposal is to demonstrate that MAPFS-Grid takes advantage of the use of a multiagent system as a conceptual framework in its design and development. The outline of this paper is as follows: Section 2 introduces the MAPFS-Grid system and describes the related work. Section 3 describes the generic structure of an agent in MAPFS-Grid. Section 4 analyzes the MAPFS-Grid cooperation model and describes the communication features of MAPFS-Grid. Section 5 shows the implementation and evaluation of MAPFS, in order to measure the influence of agents in the management of data resources. Finally, Section 6 summarizes our conclusions and suggests future work. 2. Problem statement and related work 2.1. MAPFS-Grid overview MAPFS-Grid [27,30,31] is a generic framework for increasing the performance of I/O operations in grid environments. Thus, MAPFS-Grid is a suitable approach for data-intensive applications executed on data grids. MAPFS-Grid makes use of MAPFS [26], a high-performance parallel file system for clusters of workstations. MAPFS (Multi Agent Parallel File System) has been developed in the Universidad Politécnica de Madrid since 2003. The main contribution of 0167-739X/$ – see front matter © 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.future.2008.07.011