You are not logged in.
Pages: 1
discussion about the new kernel io schedulers
4 schedulers are available to choose one of them
you can use the elevator option on boot to choose one of them, else as is chosen as default
elevator= [IOSCHED]
Format: {"as"|"cfq"|"deadline"|"noop"}
See Documentation/block/as-iosched.txt
and Documentation/block/deadline-iosched.txt for details.
"as" is default
"cfq" should be for workstations
"deadline" for some kind of databases
Offline
elevator= [IOSCHED]
Format: {"as"|"cfq"|"deadline"|"noop"}
See Documentation/block/as-iosched.txt
and Documentation/block/deadline-iosched.txt for details.
just wondering ... is there a way to get the Documentation dir without downloading the whole sources? (cvs servers? webcvs? some mirror hosting source as-is?)
The impossible missions are the only ones which succeed.
Offline
i never found a way
but i'm highly interested in a way to have access to kernel documentation
without having to download the sources
Offline
i never found a way
but i'm highly interested in a way to have access to kernel documentation
without having to download the sources
You don't have to download the whole source just the patches which are quite small.
Offline
tpowa wrote:i never found a way
but i'm highly interested in a way to have access to kernel documentation
without having to download the sourcesYou don't have to download the whole source just the patches which are quite small.
i don't have the hdd-space to keep sources of the kernel :-( .... the idea is to access the Docs only (ideal solution would be to have them somewhere online, but after searching the web i didn't find them nowhere :-( :cry: )
The impossible missions are the only ones which succeed.
Offline
btw i don't think that the patch file includes all the docs
i have always the kernel source on my hd somewhere
Offline
just wondering ... is there a way to get the Documentation dir without downloading the whole sources? (cvs servers? webcvs? some mirror hosting source as-is?)
Like this?
http://www.kernelhq.cc/browse.py?css=taichi&version=539
8)
Offline
lanrat cool link
Offline
dp wrote:just wondering ... is there a way to get the Documentation dir without downloading the whole sources? (cvs servers? webcvs? some mirror hosting source as-is?)
Like this?
http://www.kernelhq.cc/browse.py?css=taichi&version=5398)
exactly! many thanx for this link - you saved the day ;-)
The impossible missions are the only ones which succeed.
Offline
back to schedulers, does anyone already have made some experiences?
Offline
back to schedulers, does anyone already have made some experiences?
under heavy load the "deadline" makes my old machine more responsive for lots of queries / second on db's
for normal work i use the default for now (no time for experiments, while important work is not finished) (means: i use "as")
The impossible missions are the only ones which succeed.
Offline
I don't notice any difference between cfq and as.
Offline
cfqv2 gives to me more responsivity, but when I'm using it I have slow transfers from device to device (or form partition to partition). I have dvd+rw so I need good transfers on devices so I'm using as.
Gnome - The weakest link!
Linux, *not* GNU/Linux!
Offline
cfqv2 gives to me more responsivity, but when I'm using it I have slow transfers from device to device (or form partition to partition). I have dvd+rw so I need good transfers on devices so I'm using as.
You can use different schedulers on a device basis.
Here's a snip from the CK maillist:
# cat /sys/block/hdc/queue/scheduler
noop anticipatory deadline [cfq]
# echo cfq > /sys/block/hda/queue/scheduler
to use CFQ on hda.
Somebody experienced kernel panics when switching default CFQ to deadline for a specific deviice.
Con says in one of his e-mails in the thread linked below:
CFQ will be run on all drives by default, but I set my
cds/dvds to deadline and when I am doing video capture I set my hard drives to deadline and then set them back to cfq.
Complete thread @ http://bhhdoa.org.au/pipermail/ck/2004- … 01981.html
Enjoy the Linux Kernel IO Schedulers!
IceRAM
:: / my web presence
Offline
I'm having a fine time woth cfq, no problems. For me, it seems a bit quicker in some cases than as, and I haven't noticed it really falling behind anywhere.
If you develop an ear for sounds that are musical it is like developing an ego. You begin to refuse sounds that are not musical and that way cut yourself off from a good deal of experience.
- John Cage
Offline
I use the ck kernels since a while now, they are extremely impressive in terms of responsiveness. I don't think I could go back.
You should disable pre-emption if you decide to use a ck kernel. Also join the mailing list, it's full of general performance tidbits and is most interesting.
EDIT: forgot to mention, the cfq scheduler is the work of Con Kolivas, and is included in his ck kernel patchsets.
Offline
I use the ck kernels since a while now, they are extremely impressive in terms of responsiveness. I don't think I could go back.
You should disable pre-emption if you decide to use a ck kernel. Also join the mailing list, it's full of general performance tidbits and is most interesting.
EDIT: forgot to mention, the cfq scheduler is the work of Con Kolivas, and is included in his ck kernel patchsets.
I just downloaded the CK patch set for 2.6.13. You are sure that you should disable preempt? where did you learn this?
thanks
By the way, what about the "staircase" scheduler? ck's site says that it "is intrinsically interactive and responsive instead of modifying an existing design with tweaks to make it interactive" Does that mean that it is even better for a quick desktop than cfq?
It's a very deadly weapon to know what you're doing
--- William Murderface
Offline
By the way, what about the "staircase" scheduler? ck's site says that it "is intrinsically interactive and responsive instead of modifying an existing design with tweaks to make it interactive" Does that mean that it is even better for a quick desktop than cfq?
The cfq scheduler is an I/O(disc read/write) scheduler. Staircase is a cpu scheduler(default in the 2.6 is the ingosched).
As far as pre-empt is concerned I'm not sure, I leave it on but I might be wrong. The newer version of the kernel metion it's more for embedded systems.
Offline
Pages: 1