Re: [RFC][PATCH] O(1) Entitlement Based Scheduler
From: Timothy Miller (miller_at_techsource.com)
Date: 02/25/04
- Previous message: Johan van den Dorpe: "nfsd: terminating on error 104 problem"
- In reply to: John Lee: "[RFC][PATCH] O(1) Entitlement Based Scheduler"
- Next in thread: John Lee: "Re: [RFC][PATCH] O(1) Entitlement Based Scheduler"
- Reply: John Lee: "Re: [RFC][PATCH] O(1) Entitlement Based Scheduler"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Date: Wed, 25 Feb 2004 12:09:47 -0500 To: John Lee <johnl@aurema.com>
John Lee wrote:
> X Windows Performance
> =====================
>
> The X server isn't strictly an interactive process, but it does have a major
> influence on interactive response. The fact that it services a large number
> of clients means that its CPU usage rate can be quite high, and this negates
> the above mentioned favourable treatment of interactive and I/O bound
> processes.
>
> Therefore, for best interactive feel, it is recommended that the X server run
> with a nice value of at least -15. From my own testing, doing a window wiggle
> test with a make -j16 in the background and X reniced was slightly better than
> for the stock kernel.
>
> When running apps such as xmms, I recommend that they should be reniced as well
> when the background load is high. With the above setup and xmms reniced to -9,
> there were no sound skips at all (without renicing, a few skips could be
> detected).
Well, considering that X is suid root, it's okay to require that it be
run at nice -15, but how is the user without root access going to renice
xmms? Even for those who do, they're not going to want to have to
renice xmms every time they run it. Furthermore, it seems like a bad
idea to keep marking more and more programs as suid root just so that
they can boost their priority.
Not to say that your idea is bad... in fact, it may be a pipe dream to
get "flawless" interactivity without explicitly marking which programs
have to be boosted in priority. Still, Nick and Con have done a
wonderful job at getting close.
This is a tough problem.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
- Previous message: Johan van den Dorpe: "nfsd: terminating on error 104 problem"
- In reply to: John Lee: "[RFC][PATCH] O(1) Entitlement Based Scheduler"
- Next in thread: John Lee: "Re: [RFC][PATCH] O(1) Entitlement Based Scheduler"
- Reply: John Lee: "Re: [RFC][PATCH] O(1) Entitlement Based Scheduler"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Relevant Pages
|