High Performance Mining of Streaming Data Session: Bandwidth Challenge Showcase: Stretching the Bounds of Network Capacity Event Type: Bandwidth Challenge Time: 11:00am - 11:10am Speaker(s): Robert Grossman, Yunhong Gu, David Hanley, Michal Sabala, Joe Mambretti, Cees de Laat, Freek Dijkstra, Hans Bloom, Dennis Paus, Yuji Oie, Kazumi Kumazoe, Shirley Connelly, Alex Szalay Location: 611-614 Abstract: Over the past few years, the SC Community has made important advances in network protocols that can support high bandwidth data streams, even over high bandwidth delay product networks. The most important challenge today is to persist this data efficiently to disk and to process this data at line speed. In this Bandwidth Challenge, we will process multiple concurrent Gb/s data streams to: 1) compute window-based histograms, and 2) perform window-based joins on multiple streams. The former is an example of a data mining primitive, which is a core operation in data mining. The latter is a core operation for data integration. All data streams will maintain multi-Gb/s per-flow transfer rate from remote sites, including but not limited to Tokyo,, Amsterdam, Geneva, Kingston, and Chicago. The streaming data transport is supported by the UDT (UDP-based Data Transfer) protocol, a reliable UDP-based protocol that is fast, fair, and friendly. Chair/Speaker Details: Robert Grossman University of Illinois at Chicago Yunhong Gu University of Illinois at Chicago David Hanley University of Illinois at Chicago Michal Sabala University of Illinois at Chicago Joe Mambretti Northwestern University Cees de Laat University of Amsterdam Freek Dijkstra University of Amsterdam Hans Bloom University of Amsterdam Dennis Paus SURFNet Yuji Oie JGN II Kazumi Kumazoe JGN II Shirley Connelly University of lllinois at Chicago Alex Szalay John Hopkins University