Re: high system cpu load during intense disk i/o

[Dimitrios Apostolou <jimis@gmx.net>]

Hi Alan,

Alan Cox wrote:
> > > In Your oprofile output I find "acpi_pm_read" particulary interesting. 
> > > Unlike other VIA chipsets, which I know, Your doesn't use VLink to 
> > > connect northbridge to southbridge. Instead PCI bus connects these two. 
> > > As You probably know maximal PCI throughtput is 133MiB/s. In theory. In 
> > > practice probably less.
>
> acpi_pm_read is capable of disappearing into  SMM traps which will make
> it look very slow.

what is an SMM trap? I googled a bit but didn't get it...

> > about 15MB/s for both disks. When reading I get about 30MB/s again from 
> > both disks. The other disk, the small one, is mostly idle, except for 
> > writing little bits and bytes now and then. Since the problem occurs 
> > when writing, 15MB/s is just too little I think for the PCI bus.
>
> Its about right for some of the older VIA chipsets but if you are seeing
> speed loss then we need to know precisely which kernels the speed dropped
> at. Could be there is an I/O scheduling issue your system shows up or
> some kind of PCI bus contention when both disks are active at once.

I am sure throughput kept diminishing little by little with many 2.6 
releases, and that it wasn't a major regression on a specific version. 
Unfortunately I cannot backup my words with measurements from older 
kernels right now, since the system is hard to boot with such (new udev, 
new glibc). However I promise I'll test in the future (probably using 
old liveCDs) and come back then with proof.

> > I have been ignoring these performance regressions because of no 
> > stability problems until now. So could it be that I'm reaching the 
> > 20MB/s driver limit and some requests take too long to be served?
>
> Nope.

the reason I'm talking about a "software driver limit" is because I am 
sure about some facts:
- The disks can reach very high speeds (60 MB/s on other systems with udma5)
- The chipset on this specific motherboard can reach much higher 
numbers, as was measured with old kernels.
- No cable problems (have been changed), no strange dmesg output.

So what is left? Probably only the corresponding kernel module.


Thanks,
Dimitris

-
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/