After takin' a swig o' grog, Erik Funkenbusch belched out
this bit o' wisdom:
> On Wed, 11 Mar 2009 09:44:55 +0000, Roy Schestowitz wrote:
>
>> -----BEGIN PGP SIGNED MESSAGE-----
>> Hash: SHA1
>>
>> Start up performance - something that always matters
>>
>> ,----[ Quote ]
>>| Today I want to show you what we have done so far for start up performance.
>>| We made a thoroughly analysis about the start up performance under Windows
>>| and Linux. Using powerful tools (e.g. Process Monitor from Sysinternals),
>>| which provide data from the system level, made it easy to collect data to
>>| quantify the different aspects during start up. Based on this data we want to
>>| see where we can influence the start up performance.
>> `----
>>
>> http://blogs.sun.com/GullFOSS/entry/start_up_performance_something_that
>
> Wait... wait.. wait..
>
> But I thought OpenOffice.org was already blazingly fast?
>
> According to Sun's own benchmarks, it takes 24.8 seconds to start up.
>
> almost half a minute.
>
> So Terry, what was that bout OOo being so fast again?
Well, Erik, could it be <cough> <cough> the fault of the /operating system/?
The following table summarises the measured values on OpenOffice.org
Writer cold start up. Cold start means that OpenOffice.org was started
after a reboot and removing the prefetch file which created by Windows
XP. For more details about the prefetch feature of Windows XP you can
links at the end of this blog.
Aspect Time Percentage
Writer startup 24800ms 100,00%
. . .
It's obvious that file I/O plays the most important role on start up.
More than 80% of the time needed for start up is lost due to reading
libraries or data. Reading data files is about half the time
OpenOffice.org needs to read all necessary libraries. It's also clear
that raw CPU power doesn't help for a quicker start up.
Gotta love that good ol' XP I/O.
--
Being owned by someone used to be called slavery -- now it's called commitment.
|
|
