Attachment 202072 Details for Bug 352558 – Prformance report by Serge Démoulin

Some Eclipse Foundation services are deprecated, or will be soon. Please ensure you've read this important communication.

Prformance report by Serge Démoulin

performance-report.txt (text/plain), 3.24 KB, created by Ivan Furnadjiev

on 2011-08-24 06:33:33 EDT

(hide)

Description:

Filename:

MIME Type:

Creator: Ivan Furnadjiev

Created: 2011-08-24 06:33:33 EDT

Size: 3.24 KB

patch

obsolete

>===================
>PERFORMANCE TEST REPORT
>===================
>
>The test application is composed of 3 view :
>View : it is the view of the Sample
>LightComboView : a view with 15 LightCombos and each LightCombo has got 25 items (see screenshot).
>ComboView : exact the same as "LightComboView" but with RWT Combo.
>Test method :
>0. Close all the browser windows, clear the cache. Restart the browser
>1.  start the application
>2. click on "LightComboView" and start the chronometer
>3. measure the time until the view is completly shown
>4. 5. same for "ComboView"
>6. restart the application without restarting the browser
>
>
>
>The test application is attached to this mail.
>
>Performance (time)
>============
>The following table is a summary of the performance measured with RAP 1.4.
>I measure the time elapsed when I click to a View until the view is displayed.
>(1) The first number is the measured time after IE new started was and the cache is empty.
>(2) For the second number is the measured time after I restarted the application without restarting IE
>
>
>Server : The application runs on my own computer under eclipse.
>IE8 test : IE8 runs on my own computer
>IE7 test : IE7 runs on another computer in the same local network.
>
>Time elapsed to show the view
>------------------------------------
>                          LightCombo                      RWT Combo
>IE8                       (1) 0,79s  (2) 0,73s            (1) 2,19s   (2) 2,48s
>IE7                       (1) 2,98s  (2) 2,0s             (1) 11,25s  (2) 14s
>
>RWT has a lot of advantages but such performances could be a problem for using it in middle complex applications like ours.
>These measurements was made on the same local network, for the real case there is a bigger network latency.
>
>I made the measurements many times and with the RWT Combo the second measure is alway significantly worse as the first. The third measure is equivalent to the second.
>For LightCombo is the opposite and it was expected because of the caching.
>
>
>Show of the drop-down list
>-------------------------------
>The reaction time of the RWT Combo is not immediate (the time elapsed to show the drop-down list) until the LightCombo show immediatly the drow-down list. Under IE it can take 1,5s
>
>
>LightCombo
>-------------
>The LightCombo is composed of three HTML elements (1 select and 2 inputs, the two input is the button and has the "down" image).
>The select is invisible (except the drop-down list) and the inputs are used for the look & feel.
>
>
>Our real application
>-----------------------
>The test application is deduced from our real application.
>In this application has one composite that has got 37 Labels, 15 Combos, 22 TextFields and 2 Groups.
>
>Time elapsed to show this composite under the same condition that the first test (local network, IE8 and Firefox on the same computer as the server, IE7 on a different computer) :
>                          LightCombo                                 RWT Combo
>Firefox 5                 (1) 1,27s  (2) 0,97s (3) 1,16s             (1) 2,18s    (2) 1,62s   (3) 1,48s
>IE8                       (1) 2,33s  (2) 2,18s (3) 1,78s             (1) 4,26s    (2) 6,67s   (3) 7,55s
>IE7                       (1) 9,52s  (2) 8s    (3) 7s                (1) 21s      (2) 31s

Actions: View

Attachments on bug 352558: 199974 | 202072 | 205008