Special Issue: Cluster 2001
Concurrency and Computation Practice and Experience · 2003
In October 2001, more than 300 delegates gathered in Newport Beach, CA, U.S.A., to participate in Cluster 2001, the International Conference on Cluster Computing.The event was the accumulation of a tremendous effort by the organizers, in the face of a number of adverse circumstances and conditions, and resulted in an overwhelmingly successful occasion.The credit for this success not only goes to the organizers, but to the contributors, exhibitors and delegates who created a diverse and stimulating environment for those from academic, business and commercial backgrounds.The quality of the paper submissions led the technical programme committee to choose a short list of what they considered the best papers, which the authors were then asked to extend and resubmit for possible inclusion in this special issue.The resubmitted papers were peer reviewed again and, in some cases, radically updated before being finally accepted.The topics that the papers in this special issue cover are: using multiple independent networks to overcome inter-node bandwidth limitations; the design of an intelligent parallel I/O system for clusters; a load balanced Java-based cluster Web server; tools for the management and installation of clusters; the integration of a sockets interface onto a VIA; and the production of an intelligent network interface for dynamically reconfigurable clusters.As one can see from the diverse range of topics covered in this special issue, there is still a wide range of technical research areas being studied in order to fulfil the needs of the diverse range of applications that are now regularly using cluster platforms.It is clear from the talks and discussions that occurred at Cluster 2001 that cluster computing is a vibrant and dynamic area.The focus of research is moving away from the original goals of providing a basic computational platform to that of providing sophisticated and high-level tools and environments for cluster administrators and application developers.Even though cluster-related technologies are maturing, numerous research issues still need to be addressed, including heterogeneity, scheduling in time and space, check pointing and task migration, parallel debugging and performance optimization, and the development of conventions and standards for programming SMP clusters.