Performance testing (pass 1)

[Owen Taylor]

On Fri, 2003-09-26 at 10:10, Stephan Schutter wrote:
> I compared Windows XP and RHL 9 on a Dell Laptop (C610) 
> 
> CPU:	1000MHZ
> RAM:	256MB
> 
> Software: 
> 
> OS:	WinXP			RHL 9
> 	Open office		OpenOffice
> 	MS Office 2k		MS Office 2k
> 	
> 
> Software startup times:
> 		WinXP		RHL 9
> 		  (s)		 (s)
> OS bootup	37		70	
> Login		8		21
> Def Browser	1		7
> Word 2k		4		6
> Open Office	10		21
> Def TextEdit	1		3
> File Browser	1.5		1.5
> 
> Timings are average of 5 launches. Word on linux, suprisingly, is snappier 
> than OpenOffice... Open Office is faster on Windows than linux... 
> 
> Soooo.... do I need to ask?
> 
> 1. If Open Source writes better software; then why is it fatter and 
> slower?

Speed and size are really just one aspect of a piece of software.

If you want software to be small / go fast bad enough, you can
usually achieve that. With a cost in:
 
 - Time spent on the task
 - Sacrificed features
 - Maintainability

The question really should never be "how fast is the software" but
"is it fast enough". Faster is better, of course, all thngs 
being equal. But you can't forget the cost.

It's always easier to make well written software run fast than to take
badly written fast software and make it well written.

I think there is a lot of things that can be done to improve 
desktop performance on Linux. Since it's Open Source, there are also
lots of people to do that.

> 2. Is any one at RHL working on desktop performance?

At RHL? RHL is a product, not a place. :-)

Desktop performance is an *interest* of a number of people at
Red Hat, including myself. Most of whom, however, are pretty heavily
scheduled elsewhere. We certainly don't have a dedicated team of
people on that task.

One thing I hope to do over the next several months is to write
up a project list for experiments and measurements relating to
desktop performance and get that up on the Fedora web page.

> These are some of the sugestione that I may need to answe to when I 
> present this and further data to my managers. Gys... what should I say?

Well, if OS and app startup time are your only criteria, then you
have clear data for those two particular operating versions.

I would imagine that things like:

 - Cost
 - Control
 - Security
 - Stability
 - Features
 - Usability
 - Performance when running (how often do you reboot a day?)

Would also figure into the picture.

> One of the original sugestions of why to use Linux is that it could buy 
> you longer life on the old hardware... this does not seem to be the case. 

I don't think it's ever been a terribly realistic claim for office
desktop use. It does tend to be true for server applications ...
a lot of servers can be run on basically nothing for hardware.

It also can be true for "Kiosk" type applications... some of the
alternative desktops (Xfce, say) can be quite light. But if you
need to run OpenOffice, then the savings there are probably
irrelevant.

The main reason Linux is good for old/slow hardware is the control
and flexibility it gives people doing configuration, not that
we have some magic "do more with less" bullet.

Regards,
						Owen