1 / 14

Multiscale Traffic Processing Techniques for Network Inference and Control

Multiscale Traffic Processing Techniques for Network Inference and Control. Richard Baraniuk, Edward Knightly, Robert Nowak, Rolf Riedi Rice University July 2000. Rice Networking Research. INCITE (RB, EK, RN, RR, Coates) Scalable QoS (EK) Multi-tier (Aazhang, Wallach, RB, EK, RR)

gitano
Download Presentation

Multiscale Traffic Processing Techniques for Network Inference and Control

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Multiscale Traffic Processing Techniques for Network Inference and Control Richard Baraniuk, Edward Knightly, Robert Nowak, Rolf Riedi Rice University July 2000

  2. Rice Networking Research • INCITE (RB, EK, RN, RR, Coates) • Scalable QoS (EK) • Multi-tier (Aazhang, Wallach, RB, EK, RR) • ScalaServer (Druschel, Zwaenepoel) • Mobile IP (Dave Johnson) Rice University INCITE Project, July 2000

  3. Technical Challenges • State of network is intractable on a per-flow basis • Poor understanding of the origins of complex network dynamics • Lack of adequate modeling frameworks for network dynamics Manageable, reduced complexity model with known accuracy Rice University INCITE Project, July 2000

  4. INCITE InterNet Control and Inference Tools at the Edge • Overarching Objective • edge-based network measurement • modeling, monitoring, inference and control • scalable, real-time, online algorithms • (www.ece.rice.edu/INCITE) • Current DARPA Project Goals • novel traffic models:realistic, manageable • capture multiscale variability and burstiness • provide basis for a novel queuing approach and a intelligent probing strategy • synthesis and inference Rice University INCITE Project, July 2000

  5. Multiscale Nature of Traffic • Multifractal (Riedi et al. ’97) • small time scale • Network, protocol layer • Control at Connection level • LRD (Willinger et al. ‘93) • Large times • Client behavior • Bandwidth over Buffer _________ _________ Rice University INCITE Project, July 2000

  6. Multiscale Modeling Time  Scale | | \/ Innovative synthesis: multiplicative Rice University INCITE Project, July 2000

  7. Modeling on all Time Scales additive multiplicative Matching variances on all scales Positive, bursty Gaussian, LRD Rice University INCITE Project, July 2000

  8. Matching of Marginals Real Trace Multiplicative Models: Additive Models: match marginals closely match only variance 6ms 6ms 12ms 12ms 24ms 24ms

  9. MultiScale Queuing approach Queue-length = supr(Kr- rc) Kr = aggregate arrival in r time unit difficulties: non-linearity & correlated events MSQ key insight (SigMetrics, InfoComm) For MWM – traffic: overflows on dyadic times are “independent” Rice University INCITE Project, July 2000

  10. Multiscale Queuing MSQ formula: for all scales (non-asymptotic) predictive capability revolutionary queuing approach Rice University INCITE Project, July 2000

  11. Cross-traffic: Probing at Edge Abstraction of connection: multiscale statistical model of delay and loss Chirps of Probes: meet key protocol timing maximize inference capability MSQ: from queuing delayinfer cross-traffic -> improved control Rice University INCITE Project, July 2000

  12. Multifractals: A Hand on Bursts • Multifractals • Classify burstiness (quantitative and qualitative) • Captures non-Gaussianity • Multifractal models: parsimonious, tractable & realistic • New understanding • Novel statistical tools Rice University INCITE Project, July 2000

  13. INCITE: Deliverables • Multifractal Analysis Toolbox • Wavelet based estimators with known accuracy • Traffic Synthesis Software • Rapid multifractal algorithms • Network Path Modeling Toolbox • Online Inference of competing cross-traffic Rice University INCITE Project, July 2000

  14. Challenges • Improvements of algorithm • Adaptive • Passive monitoring • Deal with loss • Effect of network conditions on accuracy of inference Impact • INCITE project has promise to transform easily deployable COTS networks into predictable, controllable, and well-understood systems www.ece.rice.edu/INCITE /DARPA

More Related