Recommendations for Web Application Performance Tests

Question

I am going to start testing the web application on the intranet. In particular, I have to determine the performance of the application.

Please, someone can suggest formal / informal standards of how I can judge application performance.

Answer 1

Use the tool to test loads and loads. If you are using Java, check out JMeter . It provides various methods to test the performance of your application. You should focus on:

• Response time . How fast the application works for regular queries. Check out the read / write usage example
• Load test . How your application works in high traffic mode. The tool will present several queries (you can configure them correctly) over a period of time.
• Stress test Can your application run for a long period of time? This test will push your application to limitations.

Start with this, if you're interested, there are other kinds of tests.

Answer 2

To test the interface, YSlow is great for getting statistics on how long your pages load from a user's perspective. It is divided into statistics for each specific HTTP request, the time it receives, etc. Get it at http://developer.yahoo.com/yslow/

Firebug, of course, is also needed. You can profile your JS explicitly or in real time by clicking the profile button. Optimize if necessary and see how long all your functions have been running. This changed the way I measure the performance of my JS code. http://getfirebug.com/js.html

Answer 3

Actually, I would think that this is the response time, but other indicators that I would look at are CPU and memory usage compared to the number of simultaneous users / processes. I would also like to check that everything works as expected under normal and then peak load. You may encounter scenarios where a higher load causes application errors due to the fact that different requests come on top of each other.

If you really want to get detailed information, you will want to run various types of stress / stress tests. You will probably want to take a look at the test test (a gradual increase in the number of users in the system over time) and the spike test (a significant number of users that everyone accesses at the same time that almost no one has accessed it before). I would also run tests against the server immediately after rebooting to see how this affects the system.

You also probably want to take a look at the concept of HEAT (Testing Application Application Application). In fact, this shows what happens when some part of the system goes offline. Does the system degrade successfully? This should be the key standard.

My very big suggestion is to establish what the system should do before testing. The main reason is accountability. Ask people to acknowledge that the system must do something, and then check to see if it is correct. This is important because people immediately see the results, and this will become the base standard for the acceptable.

Answer 4

"In particular, I have to determine the performance of the application ..."

This is a complete set of issues related to the requirements covered by the expectations of your user community for what is considered reasonable and effective. Requirements have several components

• Total response time "Under load .... The site must have a total response time less than x, y% of the time ..."
• Specific response times "Under load .... Processing a credit card takes less than z seconds, in% of the time ..."
• System capacity elements: "Under load .... CPU | Network | RAM | DISK must not exceed n% of capacity ...."
• A load profile, which is a combination of the number of users and the transactions that will be conducted, during which specific, objective measures are taken to determine system performance.

You will notice that response times and other measures are not absolute. Accepting a page from six sigma production directors, the cost of moving from one exception to the 1 millionth exception is a billion times more, and the cost of moving to zero exceptions is usually not profitable for the average organization. What is considered acceptable response time for a unique application for your organization is likely to be completely different than a proposal with a high degree of compromise, which is a public Internet entity. For highly competitive solutions, Internet response expectations tend to have a 2–3 second range where user failure is dramatically undermined. This has declined over the past decade from 8 seconds to 4 seconds and is now in the range of 2-3 seconds. Some applications, such as Facebook, capture an almost imperceptible subseasonal response time for competitive reasons. If you are looking for a tough standard, they simply don't exist.

Something that helps you understand is to read a couple of industry tests for style, form, function.

• TPC-C Database Inspection Document
• SpecWeb2009 Benchmark Design Document

Setting up a reliable set of performance tests that represents your needs is a non-trivial matter. You may want to engage a specialist to carry out this phase of your quality assurance efforts.

In your choice of tool, make sure that you get one that can

• UI Exercise
• Report your requirements
• You or your team have skills in using
• You can be trained and take part in the blessing of leadership.

Skip the misfire on any of the four elements above, and you also purchased the most expensive tool on the market and hired the most expensive company to deploy it.

Good luck