Talk:Load testing: Difference between revisions

Revision as of 06:17, 21 November 2007

Revision as of 18:17, 3 January 2009

Line 7:

The two pages should be merged since the separate "tools" section can't possibly work given the guidelines: "Promotional articles about yourself, your friends, your company or products, or articles created as part of a marketing or promotional campaign, will be deleted in accordance with our deletion policies." Any listing of "tools" would have to include products, and therefore would be categorized as "marketing". To be fair, any listing of "tools" which have to include all tools or none, otherwise you'd have to get into the criteria why some are listed others are not. ([[User:Czei|Czei]] 14:41, 15 June 2007 (UTC))

. <small>—Preceding [[Wikipedia:Signatures|unsigned]] comment added by [[Special:Contributions/121.243.192.105|121.243.192.105]] ([[User talk:121.243.192.105|talk]]) 06:15, 21 November 2007 (UTC)</small><!-- Template:UnsignedIP --> <!--Autosigned by SineBot-->

== Software Load Testing is Misleading ==

The world of software load testing is changing and this article (as well as the list of load testing tools) is out of date, incomplete, and misleading. Disclaimer: I'm the founder of a new startup, [http://browsermob.com BrowserMob], and I have a very clear conflict of interest and therefore cannot make edits to any of these pages.

This article on claims that software load testing shares very little similarity with functional testing tools, primarily because load testing operates on the protocol level whereas functional testing tools work at the GUI level. I have several concerns with this:

* This paragraph would much more informational if it explained why most load testing tools work at the protocol level. Ex: because GUI-level testing for even 1 VU consumes much more memory that is practical to use for large scale load testing.

* This is no longer true. In November 2008, I launched BrowserMob, which provides load testing via [[Selenium_(software)]], a functional testing tool. It uses Selenium for both record and playback. Therefore, this is a load testing tool which does not operate at the protocol level and instead does indeed work just like a functional testing tool.

* The technique of protocol level playback is becoming more of a problem as web applications evolve and become more complex. I think this topic needs to be discussed somewhere. I recently published an article on http://ajaxian.com that discusses [http://ajaxian.com/archives/why-load-testing-ajax-is-hard why load testing Ajax applications is hard].

BrowserMob works by leveraging cloud computing, which makes it affordable to actually use real browsers for GUI-based interaction in a load test. This technique has also been [http://developer.amazonwebservices.com/connect/thread.jspa?messageID=77249 referenced independently] by people on the Amazon EC2 forums.

I also agree that the load testing tools section seems a bit odd. Some of the tools referenced are just not accurate ([[HtmlUnit]] and Selenium are almost never used for load testing, except in the case of Selenium with BrowserMob). Also, some commercial tools are referenced while some are not, which seems to violate the neutrality of Wikipedia.

I think a better approach would be to eliminate the category page and roll it up in to this article, which a more in-depth discussion of the different approaches and techniques. I would also remove HtmlUnit and Selenium entirely and put them in the proper category. I would love to do this myself (I'm the founder of OpenQA and SeleniumHQ, so I have the expertise) but I clearly have a conflict of interest.

Any thoughts would be appreciated!

[[User:Plightbo|Plightbo]] ([[User talk:Plightbo|talk]]) 18:17, 3 January 2009 (UTC)