Testing Web Application Performance Under Load
Get a clear picture of the end-to-end performance of your web infrastructure. Dotcom-Monitor’s external Windows performance counter monitoring aggregates system metrics from servers across multiple locations. Quickly compare internal metrics with real-world website performance in one convenient dashboard.
How Web Load Testing Works
Dotcom-Monitor’s Web Load Stress Test utilizes remote agents positioned worldwide to simulate real users applying a load of web “hits” to a website. Using the online interface to Web Load Stress Test the tester can set up the specific test conditions for the website or web application to ensure the most accurate test.
After choosing either a Cycle or Time option and inputting the parameters, a precise time (usually within 5 minutes of completing the setup process) when testing will start is setup. Once the test is configured, click Submit to display the cost of the test. After clicking Agree, the stress test will be queued. To view real-time testing progress during the test, click Continue. To terminate the test click Cancel.
During the test Dotcom-Monitor’s worldwide agents equally divide the simultaneous Virtual Users and start load testing the web application. The result is a comprehensive view of the web application and its ability to manage Virtual Users.
External Web Load Stress Tests Include:
- No upfront costs
- No software to install
- Easy setup, configuration, and launch
- Accurate simulation of external users (from the users’ perspective)
- Load testing of all network components, not simply the web server
- Extensive, detailed, and easy to understand reports
Benefits of External Web Load Stress Testing
There are number of internal applications available on the market for performing a stress test. While they vary in features and price, they all share one thing – they all test from inside the network, from a computer located closely to a web server.
While an internal configuration provides some insight into the performance of an application, it is not as accurate as an external web load stress test system, especially when measuring how a website will respond under a heavy load of Internet users. An end user’s experience is dependent on several components of within a network, including components of the Internet Service Provider (ISP).
Dotcom-Monitor web load stress test agents are located worldwide, and the resulting stress test data also reflects how users experience the website in those parts of the world. For example, some worldwide locations may perform better then others during the stress test, which may suggest important adjustments are to website settings with various ISPs.
In the end, an external load test stress test is a more reliable simulation of an Internet user’s interactions with a website or web application than an internal stress test and provides a more accurate report on a user’s experience.
Stress Testing Options
This type of stress test is the simplest option and is often used to determine the maximum load a website can handle before an unacceptable level of time-out messages and error codes occur. It specifies the total number of virtual users and the number of times that each virtual user will go through a transaction.
Cycle parameters include:
- the number of simultaneous virtual users
- the number of steps in the application
- the number of times each virtual user cycles through an application.
Cycle Option Example:
A test example would be a shopping-cart transaction that has seven steps and is being completed by 200 simultaneous users, where each user completes the transaction 100 times. In this example, Dotcom-Monitor agents execute the transaction 20,000 times (200 users x 100 cycles), and since it is a seven-step transaction, 140,000 Web server requests will be generated. When the cycle option is executed, there are no delays between hits, so Dotcom-Monitor puts the maximum load at any given time to your target environment.
The repetition of the shopping transaction provides an excellent gauge of whether the application will work correctly under a 200 user load.
Often used to test a website under changing conditions similar to a real-life, a Time Option stress test setup is similar to an actual user web load scenario because it provides for more flexibility in its setup. A Time Option allows for the set up of several discrete testing steps, which can each have variations, during the stress test. Those variations within each step can include the number of virtual users (the load), the number of minutes to run under such a user load, and the minimum and maximum time delays between user hits.
Time parameters include:
- number of virtual users
- number of minutes to run under such a load
- minimum and maximum time delays between hits to the website
Time Option Example:
Using the seven-step shopping cart scenario above, with the knowledge that actual users usually spend on average between three to seven seconds between clicks and that the application should handle approximately 100 simultaneous users. To measure the application response times under different user loads, the following test can be executed using the time option:
In this example, the stress test will run for 20 minutes. For the first five minutes, it will run with 50 users generating random delays between three to seven seconds. This should establish a base line for the application. Then, the test will scale up to 100 users with the same delays. This is the target test. From here, deviations can be set up to determine how the application performance changes under different conditions. For example, note that in Step 3, the user’s time decreased; and in Step 5, there was a maximum load of 200 users without any delays.