The first edition of this workshop (MGC'2003) was held on June 17th 2003, in conjunction with the 2003 Middleware Conference, in Rio de Janeiro, Brazil. MGC'2003 saw a good combination of the many different flavors in Middleware for Grid Computing, with 16 technical presentations and discussions on several stimulating topics including, among others, Classic Grids, object oriented technologies, Service-based Grids, Open grid Service Architecture (OGSA), Agent Grid, Interactive Grid, Grid Economy / Scheduling, and Portlets. The slides presented at the workshop were published on the web site as well as in the papers in the workshop proceedings. After the event, extended and thoroughly revised versions of the papers were invited to a Special Issue of Concurrency and Computation: Practice and Experience Journal (Vol.16 N.5 2004). The first workshop generated substantial interest in the community and we hope to build on this tradition this year. Far from exhausting the topics of interest, MGC'2003 paved the way for a second edition of the workshop, that will continue the most interesting and stimulating topics emerged last year, and also include some novel issues.

As network performance has outpaced computational power and storage capacity, a new paradigm has evolved to enable the sharing and coordinated use of geographically distributed resources, popularly known as "Grid" computing. The aims are to couple distributed resources and offer consistent and inexpensive access to resources irrespective of their physical location. Grid computing provides scalable and secure remote access to computing, data, and other resources throughout the Internet. These technologies enable the clustering of a wide variety of geographically distributed resources, i.e., supercomputers, storage systems, data sources, and special devices and services, that can then be used as a unified resource.

Much of the focus of "Grid" until recently has been within research organizations. Currently the involvement of commercial organization has increased tremendously. That leads to a greater focus on integration with commercial computing models and more work on the security and resource management models required for automated and secure access negotiation for remote resources. In order to control such a broad base of resources, GRID Computing has a middleware layer which controls the distributed execution of applications.