Adaptive Random Pocket Sampling for Traffic Load Measurement

트래픽 부하측정을 위한 적응성 있는 랜덤 패킷 샘플링 기법

  • ;
  • ;
  • Zhi-Li Zhang (Department of Computer Science, University of Minnesota at Twin City)
  • Published : 2003.11.01

Abstract

Exactly measuring traffic load is the basis for efficient traffic engineering. However, precise traffic measurement involves inspecting every packet traversing a lint resulting in significant overhead on routers with high-speed links. Sampling techniques are proposed as an alternative way to reduce the measurement overhead. But, since sampling inevitably accompany with error, there should be a way to control, or at least limit, the error for traffic engineering applications to work correctly. In this paper, we address the problem of bounding sampling error within a pre-specified tolerance level. We derive a relationship between the number of samples, the accuracy of estimation and the squared coefficient of variation of packet size distribution. Based on this relationship, we propose an adaptive random sampling technique that determines the minimum sampling probability adaptively according to traffic dynamics. Using real network traffic traces, we show that the proposed adaptive random sampling technique indeed produces the desired accuracy, while also yielding significant reduction in the amount of traffic samples.

트래픽 부하 측정은 네트웍 트래픽 엔지니어링의 기반이 된다. 그러나 고속 링크에서 트래픽 부하 정보를 얻기 위해 모든 패킷을 측정하는 것은, 라우터의 패킷 포워딩 성능을 저해시키므로 확장성이 결여된다. 이에 따라 샘플링 기법이 트래픽 측정의 대안으로 제시되었다. 샘플링은 라우터의 성능 저해를 최소화시킬 수 있으나 샘플링으로 예측되는 트래픽 부하는 실제 트래픽 부하와 차이를 보이게 되며, 이와 같은 오류가 제한되지 못한다면 측정값을 기반으로 하는 응용들에 부영향을 미치게 된다. 본 논문에서는 샘플링 오류를 오류 허용범위 내로 제한시킬 수 있는 적응성 있는 패킷 샘플링 기법을 제안한다. 제안 기법은 수학적 분석을 통해 얻어진 부하 예측 오류에 영향을 미치는 주요 트래픽 파라메터를 각 블록의 시작마다 예측하여 샘플링 확률을 동적으로 적응시킨다. 본 논문에서는 또한 실제 측정된 인터넷 트래픽을 이용하여 제안 기법의 확장성과 성능을 검증하였다

Keywords

References

  1. Kimberly C. Claffy, George C. Polyzos and Hans-Wemer Braun, 'Application of sampling methodologies to network traffic characterization' , in Proceedings ACM SIGCOMM '93, San Francisco, CA, September 13-17, 1993
  2. Nick G. Duffield and Matthias Grossglauser, ' Trajectory sampling for direct traffic observation' , Proceedings of ACM SIGCOMM 2000 pp271-28
  3. Nick Duffield, Carsten Lund, and Mikkel ThomP, 'Charging from Sampled Network Usage' , ACM SIGCOMM Internet Measurement WorkshoP, San Francisco, CA , Nov.2001
  4. Cristian Estan and George Varghese, 'NewDirections in Traffic Measurement and Accounting' , ACM SIGCOMM Internet Measurement WorkshoP, San Francisco, CA , Nov.2001
  5. PMA Traces Archive http://moat.nlanr.net utilization
  6. Auckland Traffic Traces http://Pma.n1anr.net/Traces/long/
  7. Donald A. Berry and Bemard W. Lindgren, Statistics theory and Methods, 2nd ed., Dux bury Press, ITP, 1996
  8. Sampled NetFlow. http://www.cisco.corn/univercd/cc/tc/doc/Product/software/ios120/120newft
  9. John M. Gottman, 'Time-series analysis', Cambridge University Press, 1981
  10. John M. Gottman, 'Adaptive Random Sampling for Load Change Detection' , Department of Computer Science and Engineering at University of Miimesota, Tectmical Report Nov. 2001
  11. A.K. Bera and C.M. Jarque, 'An efficient large-sample test for normality of observatio ns and regression residuals' , Working Papers in Economics and Econometrics, 40, Australian National University, 1981
  12. P. Billingsley, 'Convergence of ProbabilityMeasures' , New York Wisley, 1968 (P.369)
  13. C. R. Rao, 'Sampling Techniques' 2nd ed.,N.Y., Wiley. 1973
  14. W. Richard Stevens, 'TCP/IP Illustrated, vo1. 1' , Addison-Wesley, 1994
  15. http://www.caida.org/tools/measurement/coralreef/doc/doc/index.html
  16. Daniel W. McRobb, 'cflowd Design' , http://www.caida.org/too1s/measurement/cf1owd/
  17. K Claffy, Greg Miller, and Kevin Thompson, 'the nature of the beast: recent trafficmeasurements from an hitemet backbone', INET' 98 Geneva, Switzerland, 1998
  18. C. R. Rao, 'Sampliiig Techniques' , 2nd Ed., N.Y., Wiley 1973