I would like to thank the program committee for their hard work in reviewing the papers and selecting the material that appears in this conference. Each member of the technical program committee reviewed four to five papers and provided valuable feedback to the authors.
The technical program is organized in six sessions. These sessions cover recent advances in interconnects, associative technologies, networks and clusters, packet processors, optical networks, and hardware accelerators.
The conference begins with a session on recent advances in interconnect technology. Papers in this session address challenges for the design of high-bandwidth single chip networks, power management, and the evolution of traditional computer busses.
The next session addresses associative technologies for network processing. The session includes a paper on a rule grouping in TCAMs, a matching circuit for 3-stage Clos networks, and a system that provides deep packet inspection using parallel Bloom filters.
The third session explores interconnects and caching as a key technologies for networks and clusters. This session includes two papers on cluster interconnects: a scalable communication system for the ASCI Q machine and micro-benchmark level performance comparison of high-speed cluster interconnects. The session also includes a paper on buffer-caches and their impact on performance. In the evening following this session there will be a panel discussion on the role of high-performance interconnects for cluster computing.
The second day of the conference begins with a session on packet processors. The talks include a prototype exploration of the use of multiple processing elements within a network processing system, the prototype being implemented with Intel Xeon processors. Another talk discusses how network processors, such as the Intel IXP-series, can be used as building blocks in overlay networks. The third paper describes an architecture for scanning large numbers of TCP/IP traffic flows with reconfigurable hardware.
The fifth session describes recent advance in optical networks. The papers include the use slotted Wavelength Division Multiplexing to implement a bufferless, all-optical network, efficient transfer of optical bursts using a shuffle-exchange network and deflection routing, and the dynamic scheduling of optical bursts.
The final session of the conference describes hardware accelerators and network interfaces. The session begins with a paper on the end-to-end performance evaluation of a 10 Gigabit ethernet. Next, a streaming content module is presented that performs search-and-replace functions on streaming content passing through an Internet firewall. Lastly, an efficient exploitation of kernel access to Infiniband is described, and its use in implementing a distributed shared memory system.
Bryan Lyles