Special Issue on

Design and Performance of Networks for Super-, Cluster-, and Grid-Computing

to appear in the

Journal of Parallel and Distributed Computing

The need for high computation power and advances in VLSI and communications technology has led to the rapid development of cluster and grid supercomputers. Besides the technological advances, parallelism in all level of computation (thread, instruction, task, and program) has been exploited in such high-performance network-based computing systems. Almost all of such systems have their performance mainly dependent to the employed interconnection networks for inter-node communication. Many factors affect the efficiency and usability of such networks, including the topology, the switching method, the routing mechanism, and the traffic load shape. This special issue will cover various topics on networks in high performance computing systems. The topics of interest include (but are not limited to):

• Switching techniques
• Routing algorithms: unicast, multicast, broadcast, gossip, fault-tolerant, &
• Parallel and distributed algorithms and applications
• Resource placement, scheduling, and load balancing
• Topological properties, embedding, and VLSI layout of networks
• Novel network topologies
• Optical/Opto-electronic networks for high-performance system
• Mobile distributed systems
• Integrated design tool environments
• Innovative parallel simulation techniques and tools
• Mathematical performance modelling of network-based systems
• Performance analysis and evaluation
• Workload characterization and traffic modelling
• Performance-directed design of networks
• The impact of implementation technology on the design interconnection networks
• Fault-tolerant architectures
• Networks in DSM systems
• Cache-memory support
• I/O Support
• Networks in SIMD and array processors
• Networks in pipelined and vector architectures
• Case studies studying networks in successful machines
• Surveys on different issues in the design, implementation and simulation of high-performance computing systems

Prospective authors are invited to submit their original and previously unpublished papers to any one of the guest editors by May 1, 2004 . The authors are encouraged to submit by e-mail an electronic copy of the paper in pdf or ps format. If electronic submission is not possible, six hard copies of the manuscript should be submitted. Contributed papers may not exceed 20 double-spaced pages including all figures and illustrations. The first page of the submission should include the title, the names and affiliations of the authors, including complete postal and e-mail addresses, telephone, and fax numbers, a 150-word abstract, and 5-7 index terms (keywords). Submitted papers will undergo the standard review procedures of the Journal of Parallel and Distributed Computing.