Sami Sharif | University of Khartoum (original) (raw)

Papers by Sami Sharif

ABSTRACT In this paper a new model the Adjusted Autoregres- sive Integrated Moving Average (AARIM... more ABSTRACT In this paper a new model the Adjusted Autoregres- sive Integrated Moving Average (AARIMA) model for predicting Internet traffic is introduced. The steps that can be used to obtain the model are detailed. Real Internet traffic traces are used to estimate and test the models. The AARIMA model is shown to capture the local self-similarity using wavelet analysis. The model was verified using Real Internet Public Domain traces. I. INTRODUCTION AND MOTIVATION The Internet is very diverse in terms of applications and technology, new applications which consume a lot of band- width and require specific Quality of Service (QoS) are being developed continuously. Internet traffic which is generated by Internet applications is diverse and exhibits self-similar behavior. To properly model Internet traffic a through under- standing of the characteristics of the traffic is needed, these characteristics can be obtained by measurement and analysis of the actual traffic. Traffic models which can predict Internet traffic accurately are needed. The objective of this paper is to present the Ad- justed Autoregressive Integrated Moving Average (AARIMA) model for the prediction of Internet traffic, the main technique used is time series analysis and modeling. In (1) it is shown that an Internet traffic series is inherently non-stationary and therefore a global accurate H exponent tends to be difficult to estimate. The AARIMA model is presented and shown to capture the deviations in long-range dependence in Internet Traffic.

This research focuses on the comparison of two commonly used receiver initiated Dynamic Load Bala... more This research focuses on the comparison of two commonly used receiver initiated Dynamic Load Balancing (DLB) Schemes. The schemes of which are the ARR (Asynchronous Round Robin) and RP (Random Polling) are comprehensively compared theoretically and practically, weighing their pros and cons. To present a case study for their application and thus comparison, an adapted version of the Parallel Depth First Search (Parallel DFS) algorithm makes use of them in solving the famous 8-puzzle problem.

Previous studies of Internet traffic have shown that a very small percentage of flows consume mos... more Previous studies of Internet traffic have shown that a very small percentage of flows consume most of the network bandwidth. It is important to take such flows into account when measuring the network traffic self-similarity. This paper proposes a traffic measurement algorithm for self similar traffic: Adaptive sampling algorithm. This algorithm uses multistage filters and adaptation mechanism to concentrate on large flows. Using analysis and simulation, this algorithm is shown to effectively reflect the bursts of self-similar traffic over multiple time scales and can improve accuracy when network flows are self-similar.

A General Regression Neural Network (iterative model) was trained and taught to predict land cove... more A General Regression Neural Network (iterative model) was trained and taught to predict land cover features obtained from remotely sensed data. Training patterns have been generated from bands 1 and 4 of the Landsat-7 ETM+ imaging system (2000). Traditional digital image processing techniques were used to create the training and test patterns as well as analyzing the network output results. The trained network was applied to new data, for the ANN Shell (i.e. other bands of Landsat-7 and bands 2, 3, and 4 acquired by Landsat-5 TM in 1996). Reasonable outputs were obtained when the shell was applied to band 1,2,3,4,5 and 7 (Landsat-7). The trained shell failed to process most of the production patterns generated from Landsat-5 image-data. The network completely failed to process all patterns generated from spectral bands covering other areas with different topography. Based on a long series of trial and errors, it is well recognized that, the most serious problems to achieve efficient trained network are not technical, but related to training and test patterns (image data). Finally, GRNN as tested, was found to be a useful new technique for image recognition and thematic mapping from remotely sensed data.

Traditional time series models such as ARIMA models have been proven to be inadequate for modelli... more Traditional time series models such as ARIMA models have been proven to be inadequate for modelling traffic exhibiting long-range dependance. In this paper we present a new model the adjusted ARIMA model for modelling long-range dependant Internet traffic. The AARIMA model is suggested to give a quick and simple way to model Internet traffic by retaining all the properties of the ARIMA models while capturing the self- similarity. We use the Box-Jenkins methodology as a frame work for our modelling procedure. We construct our model by building the best ARIMA model possible for a trace and then adding our adjustment to obtain the equivalent AARIMA model. We show that the AARIMA model shows an evident improvement over the ARIMA model using several goodness of fit criteria, our main goodness of fit criteria is the ability to capture the Hurst parameter of the original trace being modeled. The model should not underestimate the Hurst parameter and any overestimation should be less than or equal to 20% of H parameter of the measured trace. The adjusted ARIMA model is shown to accurately predict Internet traffic for up to one hour in advance. The adjustment we propose to the ARIMA model is by introducing a feedback term made up of the first difference of the series being modeled. We used four Hurst parameter estimators to measure the self- similarity of the measured traces and both the AARIMA and ARIMA models for all measured traces. For all the estimators used the AARIMA was found to capture the long-range dependence irrespective of estimator used. We used the adjusted ARIMA model to predict three public domain internet traffic traces namely a Bellcore internet wide area network external traffic trace (length 35 hours), a Bellcore Internet Wide Area Network "purple cable" trace (length half an hour) and a MPEG-1 compressed video traffic trace (length half an hour). We show that for the public domain traces the AARIMA model gives values of H parameter which are more accura- te than those given by the ARIMA model.

Effective traffic measurement algorithms are of critical importance to properly design and implem... more Effective traffic measurement algorithms are of critical importance to properly design and implement computer networks and network services like the World Wide Web. This paper proposes a traffic measurement algorithm for self similar traffic: Adaptive multistage filters algorithm. This algorithm uses multistage filters and adaptation mechanism to concentrate on large flows. Using analysis and simulation, this algorithm is shown to effectively reflect the bursts of self-similar traffic over multiple time scales and can improve accuracy when network flows are self-similar.

Because autocorrelation functions play an important role in stochastic processes and can be used ... more Because autocorrelation functions play an important role in stochastic processes and can be used to model the traffic data practically, it is significant to study how to find a function that best fits the autocorrelation sequence of a real-traffic trace. This paper presents a asymptotic model for autocorrelation functions using Weighted non-linear least squares.

IEEE Transactions on Geoscience and Remote Sensing, Mar 1995

Chemical analysis is carried out for dust sample collected from central Sudan and the dust chemic... more Chemical analysis is carried out for dust sample collected from central Sudan and the dust chemical constituents are obtained. The mineral composition of dust are identified by the X-ray diffraction techniques. The mineral quantities are obtained by a technique developed based on the chemical analytical methods. Analyses show that quartz is the dominant mineral while the SiO2 is the dominant oxide. A simple model is derived for the dust chemical constituents. This model is used with models for predicting the mixture dielectric constant to estimate the dust dielectric constant; the results of which are seen to be in a good agreement with the measured values. The effects of the different constituents on the dust dielectric constant are studied and results are given

The study of electromagnetic signals propagation through dusty media (dust storms, air with suspe... more The study of electromagnetic signals propagation through dusty media (dust storms, air with suspended dust particles), requires knowledge of some of the electrical and mechanical properties of that media. The aim of this paper is to describe the dusty media properties related to microwave signal propagation, and to give numerical values and statistics of these parameters for dust storms that frequently occur in Sudan. Most of these values and statistics were measured and investigated by the author over the last thirty years. The parameters which are considered in this paper include (1) the dust particles geometry which is approximated by an ellipsoid, and its orientation when suspended in air (2) the dust particles "particle size distribution"; which is found to follow power law; (3) visibility statistics during dust storms as observed in some towns in Sudan, and its variation with high; (4) Permittivity and depolarization factor of the dusty medium, which is derived based on Maxwell Garnett formula and (5) dusty media Propagation Constant and Phase Shift, which are needed to estimate signal attenuation and cross depolarization.

IEEE Transactions on Geoscience and Remote Sensing, Mar 1995