bundles/org.eclipse.test.performance.ui/html/components.html - gerrit/platform/eclipse.platform.releng.buildtools - Git at Google

 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN">

 <html>
 <head>
     <meta http-equiv="Content-Type" content="text/html; charset=iso-8859-1">
     <title>Components view</title>
 </head>

 <body>
 <h1>Components view</h1>
 <p>
 This view shows the performance results in similar way than the one used to generate
 the performance results, hence make it easy to match the numbers in the corresponding
 page HTML page.
 </p><p>
 When starting the tool for the first time, this view is empty as no data has been
 populated, neither from the <a href="local_data.html">local data</a> files
 nor from the performance results database.
 </p>


 <h2>Hierarchical tree</h2>
 <p>
 Typically, the Eclipse builder runs performance tests for each component after
 the build is made on several performance test machines. Each component defines one
 or several specific performance test suites made of several test (aka scenario).
 </p><p>
 Several performance numbers (e.g. Elapsed Process Time and CPU Time) are stored
 for each scenario and all build results are available in the performance results
 database.
 </p><p>
 Hence the tree structure is made as follow:
 <pre>
 Component
    + Scenario
       + Test machine
          + Build
             + Performance numbers
 </pre>
 and may look as follow:
 <p><img src="images/components.png" alt="Components view"/></p>

 <h2>Icons</h2>
 <p>
 Several icons are displayed on tree element, here are their meaning.
 </p><p>
 The red cross means that there's at least one scenario on one machine for
 the last build with a failure (i.e. a regression over 10%).
 </p><p>
 The warning icon means that some warnings occur for some results.
 The current possible warning are:</p>
 <ul>
 <li>error over the 3% threshold on test(s)</li>
 <li>unreliable test(s): the deviation through the test(s) history is over 20%</li>
 <li>unstable test(s): the deviation through the test(s) history is between 10 and 20%</li>
 <li>no baseline for test(s)</li>
 <li>only one run on test(s)</li>
 </ul>
 <p>
 The information icon gives some other interesting information:</p>
 <ul>
 <li>the Student T-test fails on test(s)</li>
 <li>the test(s) value or its delta is less than 100ms</li>
 </ul>
 <p>
 Note that for component and scenario level, the status is the aggregation of
 the children status. That means that as soon as one scenario is in error then
 the component is also flagged in error. And of course the higher severity is
 displayed masking lower possible icons.
 </p>

 <h2>Filters</h2>
 <p>
 There are several possible filters in this view:

 <h3>Builds filters</h3>
 <ul>
 <li>Baseline: hide the baselines (starting with R-3.x)</li>
 <li>Old: hide all builds before last milestone except earlier milestones</li>
 </ul>

 <h3>Scenarios filter</h3>
 <ul>
 <li>Advanced scenarios: hide the scenarios which are not in the fingerprints</li>
 </ul>
 <p>
 As baselines results are not really useful for the survey, the filter is activated
 by default in this view. Currently the survey only concerns the fingerprint
 scenario, hence the corresponding filter is also activated by default.
 </p>

 <h2><a name="writestatus">Write status</a></h2>
 <p>
 From this view, it is also possible to write the status file for the last
 active build (see <a href="preferences.html#lastbuild">Last build</a>) by
 using the <b>Write status</b> item of the View menu:
 </p>
 <p><img src="images/write-status-menu.png" alt="Write status menu item"/></p>
 <p>
 The written status file will contain all scenarios which have failures,
 except those excluded by the status preferences set in the preferences page
 (see <a href="preferences.html#status">Status</a>)
 </p><p>
 Then it's easy to see the new regression occurring in a build when comparing
 to the previous build status file:
 </p>
 <p><img src="images/write-status-comparison.png" alt="Compare write status files"/></p>
 </body>
 </html>
	<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN">

	<html>
	<head>
	<meta http-equiv="Content-Type" content="text/html; charset=iso-8859-1">
	<title>Components view</title>
	</head>

	<body>
	<h1>Components view</h1>
	<p>
	This view shows the performance results in similar way than the one used to generate
	the performance results, hence make it easy to match the numbers in the corresponding
	page HTML page.
	</p><p>
	When starting the tool for the first time, this view is empty as no data has been
	populated, neither from the <a href="local_data.html">local data</a> files
	nor from the performance results database.
	</p>


	<h2>Hierarchical tree</h2>
	<p>
	Typically, the Eclipse builder runs performance tests for each component after
	the build is made on several performance test machines. Each component defines one
	or several specific performance test suites made of several test (aka scenario).
	</p><p>
	Several performance numbers (e.g. Elapsed Process Time and CPU Time) are stored
	for each scenario and all build results are available in the performance results
	database.
	</p><p>
	Hence the tree structure is made as follow:
	<pre>
	Component
	+ Scenario
	+ Test machine
	+ Build
	+ Performance numbers
	</pre>
	and may look as follow:
	<p><img src="images/components.png" alt="Components view"/></p>

	<h2>Icons</h2>
	<p>
	Several icons are displayed on tree element, here are their meaning.
	</p><p>
	The red cross means that there's at least one scenario on one machine for
	the last build with a failure (i.e. a regression over 10%).
	</p><p>
	The warning icon means that some warnings occur for some results.
	The current possible warning are:</p>
	<ul>
	<li>error over the 3% threshold on test(s)</li>
	<li>unreliable test(s): the deviation through the test(s) history is over 20%</li>
	<li>unstable test(s): the deviation through the test(s) history is between 10 and 20%</li>
	<li>no baseline for test(s)</li>
	<li>only one run on test(s)</li>
	</ul>
	<p>
	The information icon gives some other interesting information:</p>
	<ul>
	<li>the Student T-test fails on test(s)</li>
	<li>the test(s) value or its delta is less than 100ms</li>
	</ul>
	<p>
	Note that for component and scenario level, the status is the aggregation of
	the children status. That means that as soon as one scenario is in error then
	the component is also flagged in error. And of course the higher severity is
	displayed masking lower possible icons.
	</p>

	<h2>Filters</h2>
	<p>
	There are several possible filters in this view:

	<h3>Builds filters</h3>
	<ul>
	<li>Baseline: hide the baselines (starting with R-3.x)</li>
	<li>Old: hide all builds before last milestone except earlier milestones</li>
	</ul>

	<h3>Scenarios filter</h3>
	<ul>
	<li>Advanced scenarios: hide the scenarios which are not in the fingerprints</li>
	</ul>
	<p>
	As baselines results are not really useful for the survey, the filter is activated
	by default in this view. Currently the survey only concerns the fingerprint
	scenario, hence the corresponding filter is also activated by default.
	</p>

	<h2><a name="writestatus">Write status</a></h2>
	<p>
	From this view, it is also possible to write the status file for the last
	active build (see <a href="preferences.html#lastbuild">Last build</a>) by
	using the <b>Write status</b> item of the View menu:
	</p>
	<p><img src="images/write-status-menu.png" alt="Write status menu item"/></p>
	<p>
	The written status file will contain all scenarios which have failures,
	except those excluded by the status preferences set in the preferences page
	(see <a href="preferences.html#status">Status</a>)
	</p><p>
	Then it's easy to see the new regression occurring in a build when comparing
	to the previous build status file:
	</p>
	<p><img src="images/write-status-comparison.png" alt="Compare write status files"/></p>
	</body>
	</html>