Re: [OT]Re: Multi-tasking
El Jueves, 29 de Junio de 2006 20:31, Raúl Sánchez Siles escribió: > I currently use the deadline scheduler, do you think using another > schedule would improve things¿? Staircase scheduler rocks all the way, and also CFQ2 for I/O. But they're only available in CK patch i use here :) -- Rafael Rodríguez http://unrincon.blogspot.com
Re: [OT]Re: Multi-tasking
First of all, thank you very much for your answer. El Jueves, 29 de Junio de 2006 10:20, Dietz Proepper escribió: > Raúl Sánchez Siles: > > > In this situation, I have tried doing top on a konsole, and I find > > that the field "wa" in the above part of the top report, in the middle > > of "id" and "hi" raises to 80-90%. I don't know what exactly this field > > means, but I bet for cpu-wait state as it is the case in disk I/O. > > It's the amount of time, your systems spends waiting for i/o requests to > finish, and its quite normal to have 90%+ i/o wait while you're doing disk > i/o. In an ideal world, the disk operation should not influence the rest > of the system too much (apart from slowing down other disk i/o.) > I currently use the deadline scheduler, do you think using another schedule would improve things¿? > > All of this guessing yields me to the conclusion that I'm sometimes > > wasting up to 90% of cpu time waiting or just doing nothing, I can't > > If you're not typing very, very fast (or run some seti stuff in the > background all the time), your system spends nearly 100% of it's time > doing nothing ;-). The thing here is that I don't mind "idle-ing" the CPU when I don't need it, but bloating it waiting when I do need it. ;) -- Raúl Sánchez Siles ->Proud Debian user<- Linux registered user #416098 pgpCM7y16phGw.pgp Description: PGP signature
Re: [OT]Re: Multi-tasking
Raúl Sánchez Siles: > Here is something I have always liked to asks but I hadn't had the > chance. The problem is that on my job PC, I use reiser over a LVM over a > SATA disk. When I have some disk intensive tasks/intervals, my system > turns not as responsive as I wish and as I think should be. One part of the problem is crappy PC hardware :-\. Most certainly, the SATA tranfers block the PCI bus and use up parts of your memory bandwidth while transfering data, which is a bad thing. The other part (even with non-crap) - you're putting heavy load on your disk subsystem from one process. Now, if another process tries to do disk i/o, the heavy load from the other process slows it down. > In this situation, I have tried doing top on a konsole, and I find > that the field "wa" in the above part of the top report, in the middle > of "id" and "hi" raises to 80-90%. I don't know what exactly this field > means, but I bet for cpu-wait state as it is the case in disk I/O. It's the amount of time, your systems spends waiting for i/o requests to finish, and its quite normal to have 90%+ i/o wait while you're doing disk i/o. In an ideal world, the disk operation should not influence the rest of the system too much (apart from slowing down other disk i/o.) > All of this guessing yields me to the conclusion that I'm sometimes > wasting up to 90% of cpu time waiting or just doing nothing, I can't If you're not typing very, very fast (or run some seti stuff in the background all the time), your system spends nearly 100% of it's time doing nothing ;-).
[OT]Re: Multi-tasking
El Miércoles, 28 de Junio de 2006 13:24, cobaco (aka Bart Cornelis) escribió: > On Wednesday 28 June 2006 12:30, André Wöbbeking wrote: > > On Wednesday 28 June 2006 12:15, cobaco (aka Bart Cornelis) wrote: > > > On Wednesday 28 June 2006 05:09, Matej Cepl wrote: > > > > Hi, > > > > > > Is anything trying to write/read large amounts of data when this > > > happens? if so you might want to: > > > - check the settings for your harddisk (notably DMA) > > > - also which filessystem are you using? reiserfs3 has problems > > > locking up the system under high load > > > > I sometimes have this problem and thought that it's related to SATA. But > > I'm using reiserfs3. Do you've more infos about its problems under high > > load. > > YMMV, I can't remember where I read the details, but the FAQ at [1] has: > Q: Why do things freeze on my IDE hard drive for annoying amounts of time > A: Because when large writes are scheduled all at once, reads can starve. A >fix for this is evolving; the later your ReiserFS patch, the better we >handle this. > > (it's been over a year since I was using reiserfs, this problem made my > laptop completely unusable for 10minutes+ when doing e.g. svn update on the > d-i repository, i finally 'fixed' it by switching to ext3, situation might > have improved since then) > > [1] http://www.namesys.com/faq.html Hello All: Here is something I have always liked to asks but I hadn't had the chance. The problem is that on my job PC, I use reiser over a LVM over a SATA disk. When I have some disk intensive tasks/intervals, my system turns not as responsive as I wish and as I think should be. In this situation, I have tried doing top on a konsole, and I find that the field "wa" in the above part of the top report, in the middle of "id" and "hi" raises to 80-90%. I don't know what exactly this field means, but I bet for cpu-wait state as it is the case in disk I/O. All of this guessing yields me to the conclusion that I'm sometimes wasting up to 90% of cpu time waiting or just doing nothing, I can't understand this. As I'm not sure about the SATA details (I'm used to PATA) I think I have the best configuration, even DMA, but this go on happening. On my laptop, which includes a Dell PATA disk, the supposed "wait" states doesn't rise bigger than 30-40%. What could be happening here? Is this normal?. Thanks for your attention. I know this is OT, but I would appreciate redirection to a possible source of information related to this. Cheers, -- Raúl Sánchez Siles ->Proud Debian user<- Linux registered user #416098 pgp0dAFEy0ZPy.pgp Description: PGP signature