Parallel Computer Routing and Communication

This workshop was a continuation of the PCRCW '94 workshop that focused on issues in parallel communication and routing in support of parallel processing. The workshop series provides a forum for researchers and designers to exchange ideas with respect to challenges and issues in supporting communication for high-performance parallel computing. Within the last few years we have seen the scope of interconnection network technology expand beyond traditional multiprocessor systems to include high-availability clusters and the emerging class of system area networks. New application domains are creating new requirements for interconnection network services, e.g., real-time video, on-line data mining, etc. The emergence of quality-of-service guarantees within these domains challenges existing approaches to interconnection network design. In the recent past we have seen the emphasis on low-latency software layers, the application of multicomputer interconnection technology to distributed shared-memory multiprocessors and LAN interconnects, and the shift toward the use of commodity clusters and standard components. There is a continuing evolution toward powerful and inexpensive network interfaces, and low-cost, high-speed routers and switches from commercial vendors. The goal is to address the above issues in the context of networks of workstations, multicomputers, distributed shared-memory multiprocessors, and traditional tightly-coupled multiprocessor interconnects. The PCRCW '97 workshop presented 20 regular papers and two short papers covering a range of topics dealing with modern interconnection networks. It was hosted by the Georgia Institute of Technology and sponsored by the Atlanta Chapter of the IEEE Computer Society.

Inhalt

Keynote Address.- Approaches to Quality of Service in High-Performance Networks.- Routing I.- Integrated Multi-class Routing.- Congestion Control in the Wormhole-Routed Torus With Clustering and Delayed Deflection.- Multicasting in Irregular Networks with Cut-Through Switches using Tree-Based Multidestination Worms*.- Poster Session.- CCSIMD: a Concurrent Communication and Computation Framework for SIMD Machines.- Arctic Switch Fabric.- Router and Network Architectures I.- STREAMER: Hardware Support for Smoothed Transmission of Stored Video over ATM.- Preliminary Evaluation of a Hybrid Deterministic/Adaptive Router.- HiPER-P: An Efficient, High-Performance Router for Multicomputer Interconnection Networks.- Router and Network Architectures II (Invited Presentations).- ServerNet(TM) II.- Embedded Systems Standards.- Challenges in the Design of Contemporary Routers.- Panel Session.- Panel Session.- Messaging Layer Support.- Evaluation of Communication Mechanisms in Invalidate-based Shared Memory Multiprocessors.- How Can We Design Better Networks for DSM Systems?.- Integration of U-Net into Windows/NT.- Routing II.- Distance-Based Flow Control in Wormhole Networks.- On the Use of Virtual Channels in Networks of Workstations with Irregular Topology.- Multicasting on Switch-based Irregular Networks using Multi-drop Path-based Multidestination Worms.- Power/Performance Trade-offs for Direct Networks.- Router and Network Architectures III.- ChaosLAN: Design and Implementation of a Gigabit LAN Using Chaotic Routing.- Does Time-Division Multiplexing Close the Gap Between Memory and Optical Communication Speeds?.- Deadlock Issues.- Modeling Message Blocking and Deadlock in Interconnection Networks.- On the Reduction of Deadlock Frequency by Limiting Message Injection in Wormhole Networks.