[.NET] Load testing random pages on your web site in Visual Studio 2005

June 26, 2006 - 17:27 UTC   -   Tags: testing load testing Visual Studio I was working with load tests in Visual Studio 2005, and wanted to create a load test reading random pages of the web site. The model used in the VS2005 load tests is scenario based, which basically means each user goes through a sequence of urls from start to end. This was not what I wanted, as this is a information system, not a process based system like a webshop. An information system has endless usage scenarios. I wanted to create a web test, where the user had to open som initial pages, and could from there go to a random page on the web site.

Mining the pages
First of all, I needed a set of links to use for my test. Xenu's link sleuth is a broken link tool, which works great for my purpose. I basically crawls the site, and allows you to export a report containing all urls.

Excluding unwanted entries
The Xenu-report contains both working, broken and external links. I only wanted to exclude the broken and external links, and I also wanted to remove information from the report that I had no use for. This can be done directly from code in VS2005, but I didn't want to include code for that in my test. I used some cygwin command line tools:
cat xenu_report.txt | grep -v "not found" | grep -v "skip external" | grep -v "timeout" |
grep -v "no object data" | grep -v "connection aborted" | grep -v "invalid response" |
grep "http://" | sed -r "s/([^ \t]+)\tok.*/\1/g" > urls.txtThis command removed all the unwanted information and all the broken, timed out or external links. I wanted to test my web application, not my web server, so I decided to also remove all css, javascript, images and documents (word, pdf) from the link list:
cat urls.txt | grep -v -E "\.(css|jpg|jpeg|gif|js|jar|pdf|ico)" > weblinks.txt
Coding a VS2005 web test
To create a coded web test, either add a class file to the solution, or click generate code on a working webtest. The code below shows my test class. Notice VS2005 nice "yield return" for the enumerator. Now that's a neat language feature.
namespace LoadTestProject
{
using System;
using System.Collections.Generic;
using System.Text;
using Microsoft.VisualStudio.TestTools.WebTesting;
using System.IO;

public class CustomTest : WebTest
{
private static List urlList = null;
private Random rnd;
public CustomTest()
{
this.PreAuthenticate = false;
rnd = new Random();
if (urlList == null)
{
urlList = new List();
StreamReader fileReader = new StreamReader(new FileStream("weblinks.txt", FileMode.Open));
while (!fileReader.EndOfStream)
{
string line = fileReader.ReadLine();
if (!line.Equals(""))
{
urlList.Add(line);
}
}
fileReader.Close();
}
}
public override IEnumerator GetRequestEnumerator()
{
int numRequests = rnd.Next(3, 20); //How many requests should each user do?
for (int i = 0; i < numRequests; i++)
{
int page = i;
if (i > 2)
{
//After visiting the first 3 pages, go to a random page
page = rnd.Next(0, urlList.Count);
}
WebTestRequest request = new WebTestRequest(urlList[page]);
request.ThinkTime = rnd.Next(40, 80); //How long time does the user think before clicking a new link
request.ParseDependentRequest = false; //Do not download css and images
request.FollowRedirects = false;
yield return request;
}
}
}
}
Create a load test
Create a load test and set the CustomTest as you scenario Test Mix, you're done.