1<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN">
2
3<html>
4<head>
5	<meta http-equiv="Content-Type" content="text/html; charset=iso-8859-1">
6	<title>Components view</title>
7</head>
8
9<body>
10<h1>Components view</h1>
11<p>
12This view shows the performance results in similar way than the one used to generate
13the performance results, hence make it easy to match the numbers in the corresponding
14page HTML page.
15</p><p>
16When starting the tool for the first time, this view is empty as no data has been
17populated, neither from the <a href="local_data.html">local data</a> files
18nor from the performance results database.
19</p>
20
21
22<h2>Hierarchical tree</h2>
23<p>
24Typically, the Eclipse builder runs performance tests for each component after
25the build is made on several performance test machines. Each component defines one
26or several specific performance test suites made of several test (aka scenario).
27</p><p>
28Several performance numbers (e.g. Elapsed Process Time and CPU Time) are stored
29for each scenario and all build results are available in the performance results
30database.
31</p><p>
32Hence the tree structure is made as follow:
33<pre>
34Component
35   + Scenario
36      + Test machine
37         + Build
38            + Performance numbers
39</pre>
40and may look as follow:
41<p><img src="images/components.png" alt="Components view"/></p>
42
43<h2>Icons</h2>
44<p>
45Several icons are displayed on tree element, here are their meaning.
46</p><p>
47The red cross means that there's at least one scenario on one machine for
48the last build with a failure (i.e. a regression over 10%).
49</p><p>
50The warning icon means that some warnings occur for some results. The current possible warning are:
51<ul>
52<li>error over the 3% threshold on test(s)</li>
53<li>unreliable test(s): the deviation through the test(s) history is over 20%</li>
54<li>unstable test(s): the deviation through the test(s) history is between 10 and 20%</li>
55<li>no baseline for test(s)</li>
56<li>only one run on test(s)</li>
57</ul>
58</p><p>
59The information icon gives some other interesting information:
60<ul>
61<li>the Student T-test fails on test(s)</li>
62<li>the test(s) value or its delta is less than 100ms</li>
63</ul>
64</p><p>
65Note that for component and scenario level, the status is the aggregation of
66the children status. That means that as soon as one scenario is in error then
67the component is also flagged in error. And of course the higher severity is
68displayed masking lower possible icons.
69</p>
70
71<h2>Filters</h2>
72<p>
73There are several possible filters in this view:
74
75<h3>Builds filters</h3>
76<ul>
77<li>Baseline: hide the baselines (starting with R-3.x)</li>
78<li>Nightly: hide the nightly builds (starting with 'N')</li>
79<li>Old: hide all builds before last milestone except earlier milestones</li>
80</ul>
81
82<h3>Scenarios filter</h3>
83<ul>
84<li>Advanced scenarios: hide the scenarios which are not in the fingerprints</li>
85</ul>
86<p>
87As baselines results are not really useful for the survey, the filter is activated
88by default in this view. Currently the survey only concerns the fingerprints
89scenario, hence the corresponding filter is also activated by default.
90</p>
91</body>
92</html>