Voices Of Vision

The Art of seeing what is invisible to Others

Here we provide the exact method to calculate the delay from the trace file from NS2. Delay is one of the important factors to be characterized in any network under consideration. It directly implicates on the network performance and its efficiency. What we do is that we separate the sent packets and receive packets using the grep command. Now we have to match the packets that are sent and received. Once we match then it is just the difference between the arrival times of both the packets send and received. You can either use shell program for counting which has simple grep –c or write your own code.

First I have implemented my ARIMA model which is a mathematical model in the programming language called Java. Then I went on to explore more on some of the standard mathematical tools like Mathematica, Statistica, Gretl, R and DataPlot. Since I did not have enough time to learn all of their procedures I have to find some easy way which forced me to assume the constants of the equation. I still do not know if my method is correct or not. But then my guide said if something is to be experimentally proved then it can be considered. So I tabulated all possible values and got it done.

This post is to explain the system design of my project where I attempt to build a traffic model for wireless networks. I have created a four node wireless network scenario which generates a trace file according to the traffic in the network. This part is in NS2. After this the trace file is analyzed by JAVA and ARIMA is used for prediction. After the prediction has been done we have to change the predicted files into binary for it to be integrated with NS2. The resulting trace again is analyzed by JAVA. All the complex commands and repeated commands can be put together and executed as a shell program.

NS2 simulations produces trace files that contains various fields of information like the packet arrival time, packet size in bytes, sent or received packet, control information, source and destination node, port numbers, movements in the nodes if any, coordinates etc. All these fields ultimately characterize the state of the network at any point of time. Four our manipulation it is necessary that we extract only those fields necessary for our use. For that purpose we can use shell script or perl script. I have used a combination of grep, td, awk and other similar commands in my program. Together executing as a single shell file make things simple.

As mentioned in the previous posts a traffic model is a computer program that generates traffic like the one in real time. This can be used to test various different protocols that are being designed and implemented. It can also be used where a new network in being designed and we have to check how far the network is capable of handling the real time traffic that are expected when the network is going to be actually functional. From the traffic model proposed earlier and which was implemented in NS2 we can arrive at various design ideas like increasing the height of antenna or radius of node.

In ns2 there are various traffic applications that are being modelled. Some of them are FTP, CBR. Apart from these there are exponential and Pareto sources that can be modelled into the system. In actual traffic modelling applications that are custom built by the user we are in need of a method to integrate our own predicted or forecasted traffic into the system so that the nodes can generate the traffic that we want. This is possible through trace driven traffic applications.

You must produce a trace file with two columns. One with the packet inter arrival time and the other column with the packet size in bytes. Packet inter arrival time is the time difference between the arrivals of two successive packets. After we develop that file we must change it into binary format so that it can be integrated with the ns2 simulator. Details on how to change it into binary file can be found in the internet. We must use the Traffic trace class of ns2 to enable trace driven approach. Then we must provide the name of the file that we need to integrate. Once this is done the simulator operates with our own traffic. Results can then be analysed from the trace file.

As discussed before ARIMA model is a linear time series model that can be used in predictions and forecasting of traffic values in a network. A variant of this simple ARIMA model is the seasonal ARIMA model where we inherit the idea of correlation between the trend in the history and the current trend in hand. By doing so one of the major characteristics of the network, long range dependence is being characterized. The character short range dependence is already taken care of by the simple linear ARIMA model. The proper definition of these two terms can be observed as follows. Short range dependence can be stated as the correlation between two successive samples decreases as the time increases. Similarly the long range dependence can be defined as the correlation between the successive samples increases as the time increases. So coming back to seasonal ARIMA concept we are able to characterize both of them in one single equation. For example say we are currently in the process of characterizing at the 100th second. Now we know the recent trend in the series and we got to look for similar trends in the past values. If we happen to find one that suits or at least similar to it then we can correlate the series from there.

Today's Quote




Aim not for what you are, but for what you could be

- Lucas Hellmer

Copyright 2009 - Voices Of Vision

Blogspot Theme designed by: Ray Creations, Ray Hosting.