Evaluation of rate-based transport protocols for lambda-grids

Lambda-grids are richly interconnected collections of plentiful, geographically-distributed computing and storage resources. This rich connectivity is enabled by dedicated dense wavelength division multiplexing (DWDM) optical paths. With abundant bandwidth in the center of the network (many DWDM links), contention and sharing bottlenecks move from the network core to end systems. In such networks, traditional TCP is insufficient to provide acceptable performance. We identify the key communication characteristics of this radically different network, introducing a new multipoint-to-point communication pattern for data-intensive application. We evaluate several promising rate-based data transport protocols (RBUDP, SABUL/UDT and GTP) for lambda-grids under a range of communication patterns (single stream, multiple parallel streams, converging streams, and rapid transitions). Our experiments use a range of performance metrics, including sustained throughput and loss rate, inter- and intra-protocol fairness, protocol overhead, and rate adaptation speed to flow transitions. The results provide insights into the capabilities of these three protocols and also for improvements in design and implementation of rate-based protocols.

[1]  Robert L. Grossman,et al.  Simple Available Bandwidth Utilization Library for High-Speed Wide Area Networks , 2005, The Journal of Supercomputing.

[2]  Mark H. Ellisman,et al.  Data-intensive e-science frontier research , 2003, CACM.

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

[4]  S. Low,et al.  Understanding Vegas: a duality model , 2002 .

[5]  Jean Walrand,et al.  Fair end-to-end window-based congestion control , 1998, TNET.

[6]  Steven McCanne,et al.  On improving the fairness of TCP congestion avoidance , 1998, IEEE GLOBECOM 1998 (Cat. NO. 98CH36250).

[7]  Luigi Rizzo,et al.  Dummynet: a simple approach to the evaluation of network protocols , 1997, CCRV.

[8]  Jean C. Walrand,et al.  Fair end-to-end window-based congestion control , 2000, TNET.

[9]  Sally Floyd,et al.  Connections with multiple congested gateways in packet-switched networks part 1: one-way traffic , 1991, CCRV.

[10]  Steven H. Low,et al.  Understanding TCP Vegas: a duality model , 2002 .

[11]  Phillip M. Dickens FOBS: A Lightweight Communication Protocol for Grid Computing , 2003, Euro-Par.

[12]  Dimitri P. Bertsekas,et al.  Data Networks , 1986 .

[13]  Andrew A. Chien,et al.  The OptIPuter , 2003, CACM.

[14]  Daniel A. Reed,et al.  Grids, the TeraGrid, and Beyond , 2003, Computer.

[15]  Ami Marowka,et al.  The GRID: Blueprint for a New Computing Infrastructure , 2000, Parallel Distributed Comput. Pract..

[16]  Robert L. Grossman,et al.  Teleimmersion and Visualization with the OptIPuter , 2002 .

[17]  Donald F. Towsley,et al.  Modeling TCP Reno performance: a simple model and its empirical validation , 2000, TNET.

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

[19]  Cheng Jin,et al.  FAST TCP: Motivation, Architecture, Algorithms, and Performance , 2004, INFOCOM.

[20]  Dimitri P. Bertsekas,et al.  Data Networks: Second Edition , 1992 .

[21]  Robert L. Grossman,et al.  Data integration in a bandwidth-rich world , 2003, CACM.

[22]  Thorsten von Eicken,et al.  技術解説 IEEE Computer , 1999 .

[23]  Mark Handley,et al.  Congestion control for high bandwidth-delay product networks , 2002, SIGCOMM.

[24]  Cees T. A. M. de Laat,et al.  TransLight: a global-scale LambdaGrid for e-science , 2003, CACM.

[25]  Bogdan M. Wilamowski,et al.  The Transmission Control Protocol , 2005, The Industrial Information Technology Handbook.

[26]  Ian T. Foster,et al.  Data management and transfer in high-performance computational grid environments , 2002, Parallel Comput..

[27]  Robert L. Grossman,et al.  SABUL: A High Performance Data Transfer Protocol , 2003 .

[28]  Robert L. Grossman,et al.  Transport protocols for high performance , 2003, CACM.

[29]  Andrew A. Chien,et al.  GTP: group transport protocol for lambda-Grids , 2004, IEEE International Symposium on Cluster Computing and the Grid, 2004. CCGrid 2004..

[30]  Andrew A. Chien,et al.  Distributed virtual computers (DVC): simplifying the development of high performance Grid applications , 2004, IEEE International Symposium on Cluster Computing and the Grid, 2004. CCGrid 2004..

[31]  Sally Floyd,et al.  HighSpeed TCP for Large Congestion Windows , 2003, RFC.

[32]  R. Grossman,et al.  End-to-End Congestion Control for High Performance Data Transfer , .

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

[34]  Frank Kelly,et al.  Rate control for communication networks: shadow prices, proportional fairness and stability , 1998, J. Oper. Res. Soc..

[35]  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).

[36]  Jason Leigh,et al.  Reliable Blast UDP : predictable high performance bulk data transfer , 2002, Proceedings. IEEE International Conference on Cluster Computing.

[37]  Ian Foster,et al.  The Grid 2 - Blueprint for a New Computing Infrastructure, Second Edition , 1998, The Grid 2, 2nd Edition.

[38]  Donald F. Towsley,et al.  Modeling TCP throughput: a simple model and its empirical validation , 1998, SIGCOMM '98.