WAPT is a tool to test and analyze performance of websites, web-servers and web-applications.

In the modern postindustrial epoch, the sphere of communications is getting high profile. Staying at home, you can chat with your friends from New Guinea, book a hotel in Innsbruck, settle business matters through a videoconference with your colleagues, or just pay your bills.

In this connection there exists a pressing need for qualitative web-services of greater capacity providing stable performance.

Naturally, while developing various components of the systems, there arise tasks to test their performance, reveal systems' weak spots, determine maximal capacity and choose optimal software and hardware configurations.

To resolve the tasks developers can use WAPT – a tool to test and analyze performance of websites, web-servers and web-applications.

The main idea is to emulate numerous virtual users referring to the system at definite time interval. WAPT gives scope of means to "tune" the users, taken separately or in the aggregate, as well as their references to the system.

Such an automatic approach allows saving money and time. Testing is indeed a laborious stage demanding a lot of people and time, and employment of virtual users may drastically lessen required recourses.

WAPT abounds in various tips, explanations and pieces of advice. Though, they are so well placed, that they don't hinder you from work. The program's interface, in general, is very simple and efficient. It consists of several tabs united into 2 groups: to tune and test, and to view results. The only thing that seems to be superfluous is the animation at the change of groups of tabs.

To create test scenarios there may be used a very effective wizard which lets to develop a template for basic tests, such as determining an average time of server/application response, emulation of static load, stress testing, etc.

The Recorder tab provides a simple and convenient interface to determine a set of objects under test. This is a web browser allowing remembering as an object under test the consequence of attended web pages or referred web-applications.

Though, in my opinion, the tab is quite useless as it doesn't help to set options and serves only to outline objects to test. While working with the program, to define objects under test we used another tab – Editor.

The Editor tab lets to create and tune objects to test. You can set the emulation level of connection of virtual users as well as authorization support, custom HTTP headers, etc.

The most interesting feature is the support of query parameters evaluation. There is a possibility to use both static and pseudo-dynamic data (static values, lists, sequences, random figures and strings) and dynamic once based on the previous server/application response (session variables, data dynamically generated by web-applications, etc.).

This allows to create complicated chains of queries to emulate specific user's behavior. For example, their registration in the system, entering protected zones and operating there.

The Runner tab provides options setting the number and behavior of virtual users, sequence of objects which will be enquired, characteristics which will be evaluated, etc.

Here you can also schedule a test run and choose a format and paths to save reports and log-files.

We used the WAPT program to test the performance of several search engines, such as Google, Altavista and OOKO Search. The latter is developed by our company and specifically oriented towards the search of software.

The search engines were enquired for the search by most commonly used keywords which were taken from a prearranged list. While testing, the number of virtual users increased up to 20, the maximal number for a trial version.

The test results are quite informative and can be saved in HTML and XLS formats.

They can be viewed immediately after a test has been conducted using bookmarks of the Results group.

The next graph shows that the Altavista search engine started functioning in an unpredictable way and ended up displaying an error message. The transaction time dropped till zero. One of the possible explanations may be that a mechanism protecting the engine from an automatic search triggered the cut-off.

The graph below shows a stable increase of response time under the increasing load for OOKO Search system. Nevertheless, it doesn't bring about an error. And Google demonstrates a stable performance, what is not something unusual as it systematically undergoes loads higher than that provided by a trial version of WAPT. Taking this into account, when testing OOKO Search we considered Google a standard.

The testing results testify that the OOKO Search is able to demonstrate stable performance, but quite low capacity in processing large number of requests per a unit of time in an acceptable period of time.

Let us reveal the system's breakpoint by means of stress testing. Below is a resultant graph.

As we can see, we didn't get any new results. That is, thought the response time is increasing, the system still works errorless.

So, we found out that OOKO Search performance is stable; an average time of transaction is in direct proportion to the number of users and, under the peak load, counts 2.7 sec. But the restriction to 20 virtual users doesn't allow to reveal the top capacity of a system, and makes a trial version of the program quite useless.

A flexible system of prices and good support obviously contribute to the attractiveness of WAPT for the specialist developing and optimizing web-systems of high load and large number of users.

The program is a shareware product. A trial version allows to emulate up to 20 virtual users and stop working in 30 days.

WAPT keywords: load testing, stress testing, performance testing, application testing, website testing, web load, stress test tool