Special section: Paradigms for scalable and dependable grids

Since the introduction of the concept in 1998, computational grids have attracted a vast amount of attention from researchers and potential users alike. While the former recognize grids as a vehicle for exploring new ideas and approaches in advanced distributed computing, users are allured by the promise of computational power needed to attack challenging scientific applications in areas such as physics, chemistry or biology. Progress in hardware development – both in terms of processor performance and their quantity – as well as significant research funding on many continents allowed to fulfill this vision in terms of larger and more sophisticated production and research installations. However, these trends to larger and more complex grids might bring the users more headaches instead of benefit from the vast computational power. Issues arise when the number of servers and other resources increase beyond the levels known from existing clusters and data centres. The main reasons are: inability of existing software architectures and implementations to cope with large scales; increased likelihood that at least one component is faulty, which makes a (partial) failure of a normal situation; and finally, the effort needed to maintain and manage such systems, which calls for automated approaches. The papers collected in this special section target these problems by suggesting techniques for improving scalability, dependability and resource availability in grids. In addition to these primary focus points, two of the papers discuss their results in the context of desktop grids. These collections of commodity PCs are becoming increasingly attractive by harnessing the untapped computational power of non-dedicated resources such as office or even private computers. The five papers of this special section have been selected from the 14 submissions to the 2nd CoreGRID Workshop