3:30, SPEC-L5.1
SAINT: NETWORK TOMOGRAPHY FOR INTERNAL DELAY ESTIMATION
M. COATES, R. NOWAK
On-line, spatially localized information about internal network performance can greatly assist dynamic routing algorithms and traffic transmission protocols. However, it is impractical to mea-sure network traffic at all points in the network. A promising alternative is to measure only at the edge of the network and infer internal behavior from these measurements. In this paper we concentrate on the estimation and localization of internal delays based on end-to-end delay measurements from sources to receivers. We develop an EM algorithm for computing MLEs of the internal delay distributions in cases where the network dynamics are stationary over the observation period. For time-varying cases, we propose a sequential Monte Carlo procedure capable of tracking non-stationary delay characteristics. Simulations are included to demonstrate the promise of these techniques.

3:50, SPEC-L5.2
ESTIMATING LONG-RANGE DEPENDENCE IN IMPULSIVE TRAFFIC FLOWS
X. YANG, A. PETROPULU, J. PESQUET
Traffic flow in high-speed data network systems is often impulsive and long-range dependent. Impulsiveness implies a heavy-tailed marginal distribution, thus lack of finite second-order statistics. Hence, traditional methods for quantifying the long-range dependence of traffic based on its second-order statistics. Hence, traditional methods for quantifying the long-range dependence of traffic based on its second-order statistics are not applicable. Long range dependence and self-similarity play an important role in traffic engineering. We have recently shown that the generalized codifference can quantify the dependence structure of impulsive self-similar processes, such as high-speed network traffic. In this paper, we propose an estimator for the generalized codifference and provide the conditions for it to be asymptotically consistent. We show that these conditions are satisfied for the EAFRP which is a process proposed for modeling high-speed network traffic. We provide simulations results to demonstrate the properties of the proposed estimator, and show how it can be a useful tool in maintaining fairness among users sharing limited network resources.

4:10, SPEC-L5.3
AN ADAPTIVE BROADBAND ESTIMATOR OF THE FRACTIONAL DIFFERENCING COEFFICIENT
C. HURVICH, E. MOULINES, P. SOULIER
We consider semiparametric fractional exponential (FEXP) estimators of the memory parameter d for a potentially nonstationary linear long-memory time series with smooth additive trend. We use differencing to annihilate the trend, followed by tapering to handle the potential non-invertibility of the differenced series. We propose a method of pooling the tapered periodogram which leads to more efficient estimators of d than existing pooled, tapered estimators. We establish asymptotic normality of the estimator. Finally, we consider minimax rate-optimality and feasible nearly rate-optimal estimators. Some simulations are presented to illustrate our findings. Applications to measure the Hurst coefficient of network traffic data will be presented at the time of the conference.

4:30, SPEC-L5.4
UNICAST INFERENCE OF NETWORK LINK DELAY DISTRIBUTIONS FROM EDGE MEASUREMENTS
M. SHIH, A. HERO
Inference of network internal link characteristics has become an increasingly important issue for operating and evaluating large telecommunication networks. Since it is usually impractical to directly monitor each link along a specific path, end-to-end probes are sometimes used to collect link characteristic information at edge nodes of the net-work. This paper deals with unicast probing methods for estimation of link delay characteristics. Unicast traffic is easy to generate and is supported by almost every network currently in operation. Under the assumptions that link de-lays are spatially and temporally independent, we propose a bias corrected estimator for the internal link delay cumulant generating function (CGF) based on unicast probe end-to-end delay measurements. Through simulation we show that the proposed estimator attains a level of mean squared error comparable to link delay CGF estimates obtained from directly measured link delay statistics. We can use these CGF estimates to estimate delay mean, variance and level exceedance probabilities for each link.

4:50, SPEC-L5.5
STATISTICAL SCALING ANALYSIS OF TCP/IP DATA USING CASCADES
S. ROUX, P. ABRY, D. VEITCH, L. HUANG, J. MICHEEL, P. FLANDRIN
The scaling properties of Internet data are analysed in detail through the unifying viewpoint of Infinitely Divisible Cascates. From exeptionally precise TCP/IP traffic traces are extracted time series including arrival rate, durations, and interarrival times of TCP connections. We show that IDC's offer a pertinent description of these series. Relations between them are investigated, yielding insights on the sources of the scaling and possible modelling approaches.

5:10, SPEC-L5.6
WAVELETS AND MULTIFRACTALS FOR NETWORK TRAFFIC MODELING AND INFERENCE
V. RIBEIRO, R. RIEDI, R. BARANIUK
This paper reviews the multifractal wavelet model (MWM) and its applications to network traffic modeling and inference. The discovery of the fractal nature of traffic has made new models and analysis tools for traffic essential, since classical Poisson and Markov models do not capture important fractal properties like multiscale variability and burstiness that deleteriously affect performance. Set in the framework of multiplicative cascades, the MWMprovides a link to multifractal analysis, a natural tool to characterize burstiness. The simple structure of the MWM enables fast O(N) syn-thesis of traffic for simulations and a tractable queuing anal-ysis, thus rendering it suitable for real networking applications including end-to-end path modeling.