Communication Networks Research - ensc.sfu. ljilja/cnl/presentations/laxmi/BCNET2008/IRMACS_2008...Communication

Communication Networks Research - ensc.sfu. ljilja/cnl/presentations/laxmi/BCNET2008/IRMACS_2008...Communication

  • View
    0

  • Download
    0

Embed Size (px)

Text of Communication Networks Research - ensc.sfu....

  • Communication Networks Research Laxmi Subedi, Mohamadreza Najiminaini, and Ljiljana Trajković, {lsa38, mna28, ljilja}@cs.sfu.ca Communication Networks Laboratory, School of Engineering Science, Simon Fraser University

    Communication Networks Laboratory http://www.ensc.sfu.ca/research/cnl

    Packet loss is an indicator of congestion and it causes TCP to reduce its congestion window size and throughput. TCP is designed for wired networks, where packet loss is mainly due to congestion. In wireless networks, TCP packet loss is often caused by transmission errors and handoffs. Hence, there is no need to reduce the TCP congestion window size.

    Improving TCP over wireless links

    ACK(n-1) ACK(n)

    Mobile host (receiver)

    Fixed host (sender) M-TCPSH-TCP

    TCP M-TCP

    Super host (router)

    M-TCP network architecture.

    Modeling traffic from a public safety wide area network

    Repeaters

    Consoles Dispatch

    Network Gateway

    Air Interface

    Data

    PSTN

    Data Port Management

    Other EDACS Systems

    T1

    M-RK

    M-RK

    M-RK NETWORK SWITCH

    Data Gateway

    PBX Gateway

    T1

    T1

    Other EDACS Systems

    T1

    E-Comm network architecture.

    Statistical analysis of collected data reveals that: call inter-arrival times are long-range dependent call holding times follow lognormal distribution and are uncorrelated.

    Clustering analysis is used to model user behavior. Traffic is predicted using Seasonal Autoregressive Integrated Moving Average (SARIMA) models.

    E-Comm network coverage.

    Analyzing packet loss in IP networks

    A signal is decomposed into a hierarchical set of approximations and details using wavelet analysis. The detail part provides information in the time and frequency domains. It serves as a mathematical microscope for investigating complex traffic and loss pattern.

    Loss pattern analysis: details on various time scales.

    Characterize, model, and analyze collected traffic data. Apply statistical analysis and search for traffic invariants: self- similarity and long–range dependence. Understand underlying dynamical behavior of complex systems representing packet networks.

    BCNET Technology Conference April 22 - 23, 2008

    References: • W. G. Zeng and Lj. Trajkovic, “TCP packet control for wireless networks,” Proc. IEEE Int. Conf. on Wireless and Mobile Computing, Networking and Communications (WiMob 2005), Montreal, Canada, Aug. 2005, vol. 2, pp. 196–203. • F. Xue, V. Markovski, and Lj. Trajkovic, “Wavelet analysis of packet loss in video transfers over UDP,” Proc. Internet Computing 2000, Las Vegas, NV, June 2000, pp. 427–433. • D. Sharp, N. Cackov, N. Laskovic, Q. Shao, and Lj. Trajkovic, “Analysis of public safety traffic on trunked land mobile radio systems,” IEEE J. Select. Areas Commun., vol. 22, no. 7, pp. 1197–1205, Sept. 2004. • H. Zhang, M. Liu, V. Vukadinovic, and Lj. Trajkovic, “Modeling TCP/RED: a dynamical approach,” in Complex Dynamics in Communication Networks, Lj. Kocarev and G. Vattay, Eds., Berlin: Springer, Series: Understanding Complex Systems, 2005, pp. 251–278.

    Measuring traffic from BC.NET

    10000

    Capture system

    10/100/1000/10000

    Ethernet switch 1

    10/100/1000

    Capture system

    10/100/1000

    Ethernet switch 2

    Supplier 1 Supplier 2 Supplier 3

    Proposed additions 10 GigE link 1 GigE link

    BC.NET router

    BC.NET client diagram.

Related documents