I wanted to accelerate booting. I found a message sayingOn manual attempts to start it I got the message: I looked into that. Here a MIN_TIMESLICE and MAX_TIMESLICE and other things can be defined. Okay, so far so good, but:

My system runs stable without this. What's it good for then? I googled, searched SuSE / Novell and came up with no explanation. Does it concern threading? Anybody a link or infos?

P.S.: And then there are sysctl and sysctl.conf. Don't these already allow to set these parameters as desired?

P.P.S.: What about niceing and renice?

P.P.P.S.: Are these stupid questions, i.e. did I overlook something?

Timeslice length is parameter for Linux scheduler - the part of the kernel that switches processes. Long timeslice is good if you run long calculations and you don't mind if your machine needs some time to answer when you press a key etc. Short timeslice is good if you want your system to respond fast, but it spends much time switching processes (so overall performance not as good as with long slice). If you want to chage the parameters you need to know exactly why and how to change. Bad aprameters may cause really poor performance.

Two things you can read:
http://www.samspublishing.com/articl...?p=101760&rl=1
http://www.tldp.org/HOWTO/KernelAnalysis-HOWTO-6.html

Quote:

What about niceing and renice?

You can use tehm too, but they affect the scheduler in a slightly different way. You can experiment with them in 'top'. I guess it's the best way to see how it works.

Quote:

Are these stupid questions, i.e. did I overlook something?

Not stupid, for sure. Rather advanced.

Yes, I didn't dare to touch anything, but still I wanted to know...

Thanks. This answers my question about niceing nicely .
It also gives some real values for MIN_TIMESLICE, timeslice and MAX_TIMESLICE. Nice

From this I gather, that timeslices have mixed points of reference; one being the tuneable software parameter "timeslice" to determine a new length if desired, the other the hardware like PIT 8253 (Programmable Interval Timer). How, then, do the distributions find out what kind of hardware the customers use or is there only the PIT 8253 with a clock of 1.19318 MHz "in the wild"?

*blush* Thank you, I like learning. And thanks also for your time, effort and the links .

Here you've got two meanings - second one is the time between two clock interrupts (which is constant). The parameter HZ can be changed, however, and is different in different architectures. This understanding of a timeslice is more low-level than the previous one.

Hi Mara, look at this reference, now I understand at last

http://tech.dupnica.net/understandin...rnel/ch5b.html

In case you didn't know it already it's perhaps a nice addition to your collection of bookmarks .

Thanks for all your help .

It's a from the book "Understanding the Linux Kernel" - good one, in fact. Note that, if the online text is not upgraded, it refers mostly to 2.2 and sometimes to 2.4. 2.6 made a number of improvements, so it's a good thing to see what changed.

Well, any one know how change/modify the timeslice of a thread or process during run-time ?

Thanks!