Coordination between parallel TCP streams on Long Fat Pipe Network

“Data Reservoir” is a file sharing facility system which uses parallel streams for huge data transfer on Long Fat Pipe Network. While experimenting Data Reservoir system, we observed that performances of parallel streams disperse a lot, which results in damage on total performance. To equalize throughput of parallel streams, we propose “Dulling Edges of Cooperative Parallel streams(DECP)”. This paper shows our observation, implementation of interface for sharing information between streams and external scheduler, preliminary experiments to compare algorithms, then finally, experimental results of file transfer from disks to disks on one and a half round trip from Japan to U.S., for 15,000 miles, which has 350ms RTT and 8.2Gbps bandwidth in SC2003 bandwidth challenge.

[1]  Cheng Jin,et al.  FAST TCP: Motivation, Architecture, Algorithms, Performance , 2006, IEEE/ACM Transactions on Networking.

[2]  Brian D. Noble,et al.  Improving throughput and maintaining fairness using parallel TCP , 2004, IEEE INFOCOM 2004.

[3]  S. Floyd.  HighSpeed TCP for Large Congestion Windows , 2003, RFC.

[4]  Tom Kelly,et al.  Scalable TCP: improving performance in highspeed wide area networks , 2003, CCRV.

[5]  Kei Hiraki,et al.  Data Reservoir: Utilization of Multi-Gigabit Backbone Network for Data-Intensive Research , 2002, ACM/IEEE SC 2002 Conference (SC'02).

[6]  Brian D. Noble,et al.  The Effects of Systemic Packet Loss on Aggregate TCP Flows , 2002, ACM/IEEE SC 2002 Conference (SC'02).

[7]  H. Koga,et al.  Data reservoir: a new approach to data-intensive scientific computation , 2002, Proceedings International Symposium on Parallel Architectures, Algorithms and Networks. I-SPAN'02.

[8]  Brian D. Noble,et al.  The end-to-end performance effects of parallel TCP sockets on a lossy wide-area network , 2002, Proceedings 16th International Parallel and Distributed Processing Symposium.

[9]  Robert L. Grossman,et al.  PSockets: The Case for Application-level Network Striping for Data Intensive Applications using High Speed Wide Area Networks , 2000, ACM/IEEE SC 2000 Conference (SC'00).

[10]  Mark D. Hill,et al.  The impact of data transfer and buffering alternatives on network interface design , 1998, Proceedings 1998 Fourth International Symposium on High-Performance Computer Architecture.

[11]  Janey C. Hoe,et al.  Improving the start-up behavior of a congestion control scheme for TCP , 1996, SIGCOMM '96.

[12]  Larry L. Peterson,et al.  TCP Vegas: End to End Congestion Avoidance on a Global Internet , 1995, IEEE J. Sel. Areas Commun..

[13]  John S. Heidemann,et al.  Effects of ensemble-TCP , 2000, CCRV.

[14]  S. Floyd,et al.  Tcp-friendly unicast rate-based flow control , 1997 .

[15]  Michael J. Karels,et al.  Congestion avoidance and control , 1988, SIGCOMM '88.