Performance testing (pass 1)

Bill Anderson bill at noreboots.com
Sun Sep 28 12:22:15 UTC 2003


On Fri, 2003-09-26 at 22:49, stephan schutter wrote:
> Bill Anderson wrote:
> 
> >
> >
> >Can we see a copy of /var/log/bootlog to see what all was started during
> >boot up?
> >
> >  
> >
> I will send this as soon as I get back to work... nothing special; just 
> the default setup...

There may be stuff tht should not be on in the default bootup.
Personally, I disable kudzu.I know when I add hardware. Disable kudzu
and rerun/post just the bootup times, would you pretty please?
:-D

> You know, I was thinking (warning):  why not do the rc scripts so that 
> some things occur in paralell? For example; if there are services that 
> need to start and they do not depend on networking being started, then 
> why not launch them in paralell, in the background?  Forking the startup 
> process tree and maximizing the IO bandwidth utilization, and CPU, would 
> make the bootup fast (important on laptops; because you start them and 
> stop them all the time as you need to).

There is some work on this, but it is more complex than it initially
sounds. You need a dependency tracker, and it needs to be smart enough
to detect certain config changes, or be "dumb" and less effective. For
example, does the printer daemon have network enabled, or does the the
MTA (vs. used in localhost fashion), or how about things like SWAT for
example.

But I agree, it should be better done, that's why I'm redoing it in
Python. ;)

>  
> 
> >
> >Is this runlevel 5 login, or console login and startx?
> >  
> >
> runlevel 5, after shutdown -r now...
Thanks.

> >I don't think these are apples->apples.
> >Here is why:
> >1) Default browser: are they different browsers? Mozilla is available on
> >both platforms, use that then compare.
> >2) I take it Word 2k is done in Crossover?
> >3) Open Office loads the needed guts for most of Open Office, whereas
> >W2k only opens W2k.
> >  
> >
> 1) good point; this is another app to test (that is fair)... But 
> remember, the drooling end user only knows that it is "the internet" and 
> does not distinguish between mozilla or IE. Perhaps coosing default 
> browser carefully would help here as well. But I will give you new 
> timings next week.

Excellent. There are other things that can be done for the drooluser.
Such as using galeon with the background start method. But of course,
galeon is gone from the default.

> 2) Your astuteness is only surpassed by your grace good sir! This is 
> like running a 16 bit app on NT... no complaints here. It works quite 
> smoothly (I use it daily at work(exposure testing ;=) )), performance is 
> OK.  Not bad for something running in app compatabillity mode.

I've heard this. I was curious as to whether or not it was the
codeweavers WINE or the straight WINE (would that be it's "vintage" ;) )


> 3) Shame on them!

Agreed.

> >And finally, how was the timing done. Are we talking wall clock or what?
> >  
> >
> Just stop watch. After all, we are talking about end user impressions, 
> so we do not need millisecond acuracy?

Ok.

-- 
Bill Anderson
RHCE #807302597505773
bill at noreboots.com







More information about the test mailing list