Difference between revisions of "David Manley perf comments"

From CDOT Wiki
Jump to: navigation, search
(Questions)
(Questions)
Line 30: Line 30:
 
*This should be BOLD CAPS MAJOR, almost lost all my wiki edits.
 
*This should be BOLD CAPS MAJOR, almost lost all my wiki edits.
  
==Questions==
+
==Skewing the Test==
*How much of this test is reliant on applications currently running on the system?
+
*How much of this test is reliant on applications currently running on the system?  Can one skew the test results to give inaccurate readings by overloading the system?
  
*By running a quick test I realized that by overloading the processor, the results of the test become much less useful.  During the test I ran a multitude of applications:
+
*By running a quick test I realized that by overloading the processor, the results of the test become much less useful.  Before running the test framework I started a multitude of applications to occupy the processor:
**2 x YouTube Flash Video
 
**1 x disk defragmenter
 
**1 x Flash game
 
**1 x Flash site
 
**1 x Windows Media streamed video
 
  
*By looking at the images below, the Normal test's % of processor peaks at approximately 26.5%.  In contrast the overloaded test shows a low point of 26.5%.  This result actually has nothing to do with Firefox, but everything to do with stress put on the processor from external applications.
+
**2 x YouTube Flash Video Sites
 +
**1 x Disk Defragmenter
 +
**1 x Flash Game
 +
**1 x Flash  Media Site
 +
**1 x Windows Media Streamed Video
 +
 
 +
*By looking at the images below, the Normal test's % of processor peaks at approximately 26.5%.  In contrast the overloaded test shows a low point of 26.5%.  This result actually has nothing to do with Firefox, but everything to do with stress put on the processor from external applications.  In addition the other results were affected aswell.
  
 
*Normal Test
 
*Normal Test
**[[Image:normalTest.png]]
+
[[Image:normalTest.png]]
  
 
*With Overloaded Processor
 
*With Overloaded Processor
**[[Image:overloadTest.png]]
+
[[Image:overloadTest.png]]

Revision as of 18:59, 29 November 2006

Comments for Performance Documentation

Framework Prereqs

  • Additional details regarding Framework Prerequisites would reduce confusion. There isn't any harm in "dumbing it down" a level.
    • Order of installations
    • Python options
    • Cygwin options (packages)
    • Where to put the .dll

Framework Config

  • Are given path values the defaults? Is there any forseeable reason that they would be different?
  • Confusion - you seem to be contradicting yourself
(Optional) For the page load test to work, modify the hostperm.1 file

    * Set to allow urls with scheme:file to open in new windows
    * Set the preference to open new windows in a tab should be off. 
    # Permission File
    # This is a generated file! Do not edit.

Config File

  • What do I modify? The paths?
    • Highlight sections that need to be modified or "may" need to be modified. Within code, could be hard to understand.
Running the test will close any existing Firefox windows.  
  • This should be BOLD CAPS MAJOR, almost lost all my wiki edits.

Skewing the Test

  • How much of this test is reliant on applications currently running on the system? Can one skew the test results to give inaccurate readings by overloading the system?
  • By running a quick test I realized that by overloading the processor, the results of the test become much less useful. Before running the test framework I started a multitude of applications to occupy the processor:
    • 2 x YouTube Flash Video Sites
    • 1 x Disk Defragmenter
    • 1 x Flash Game
    • 1 x Flash Media Site
    • 1 x Windows Media Streamed Video
  • By looking at the images below, the Normal test's % of processor peaks at approximately 26.5%. In contrast the overloaded test shows a low point of 26.5%. This result actually has nothing to do with Firefox, but everything to do with stress put on the processor from external applications. In addition the other results were affected aswell.
  • Normal Test

NormalTest.png

  • With Overloaded Processor

OverloadTest.png