Network Tomography For Internal Delay Estimation Using Machine Learning

In recent days Computer networks are becoming increasingly large and such complex;especially with the implementing of internet into all the thing people life(Internet of Thing or IoT).It is essential to be able to monitor and to analyze networks in a timely and efficient manner;in order to extract important metrics and measurements and to do so in a way which does not affect the performance of the network under interest.However,it is impractical to measure network traffic at all points in the network.A promising altermative is to measure only at the edge of the net.work and infer intermal behavior from these measurements.In order to solve the problem of interior links parameters measurement(e.g.delay and packet l0ss rate),this dissertation has adopted the technique of Network Tomography(NT)which is collecting the measurement of path performance based on end-to-end measurement then infer the probabilities distribution of the logical links performance using statistical calculations.This technique of estimating the links performance from end-to-end performance neither need any cooperation from the internal network,nor dependence from communication protocol.Furthermore,a new technique of statistic method is used with the network tomography to make it easier for the modeled network to improve the performance of estimating the interior links performance parameters.Which is the Machine Learning(ML)and specifically the Linear Regression.This technique has the ability to predict real values(e.g.links delay)from a give input such as paths delay by making the model learning from a given example data(Learning data which about%80 of the total data)and then testing the model with a percentage of%20 of the data(Test data)to evaluate the model.The estimated delay values is compared to the actual delay values generated from the simulation of a wired network using Network Simulator NS2.The error(Mean Square Error MSE)is calculated to evaluate the performance of the model.