Re: IDE disk slow? There's help...

From: Michael Kwasigroch (mkwasigr@intercope.com)
Date: Mon Oct 23 2000 - 04:22:39 EST


Hi Mark, hi all

thanks to all your feedback, but please allow me to add some thoughts to
all this:

I agree that in this supersonic world of PC hardware with product cycles of
4 weeks or so that my board and SCSI hardware is not the latest, best and
fastest, it may even be obsolete, but it's still stable and does what I
need and I currently don't want (and need) to invest in some new
hardware...

But you say "2.2.17 is obsolete"...?

Frankly: are you always trying to be on the bleeding edge? If 2.2.17 is
obsolete then about 99% of Linux users are using an obsolete kernel. I know
we had this dicsusion before and some people think that once a kernel is
declared stable it is obsolete and thus uninteresting but we're talking
here about Linux as a **PRODUCT** which should give 100% stability along
with appropriate performance for the majority of users.

Ok, if the ide-patch is considered to introduce new problems in the stable
kernel I agree to leave it out of 2.2.x. I also want 2.2.x to be as stable
as possible. But if people find that stock 2.2.x is not giving the
performance as some other OS, well, it will definitely not increase the
perceptance of Linux...

One other aspect of all this is that the majority of users may not even
notice that stock 2.2.x IDE performance may not be optimal because some
(most?) of the distros around tend to patch the default kernel. I (still)
use SuSE 6.4 (may be obsolete as well ;-) and it's a great product. I'm
currently not sure if they added some version of the ide-patch to their
default kernel but I don't care since I always use the latest stable stock
kernel with some (hopefully) well chosen patches. I also know that SuSE and
other distros contain stock kernels as well.

I didn't care about IDE at all and I even compiled my kernel entirely
without the IDE driver until I bought this IDE disk. As you probably
guessed by now, Andre's ide-patch joined the list of my "well chosen"
patches...

Punchline:
- I still think that my hardware is not too old
  and most possibly is still in use at some sites.

- I do __know__ that IDE performance could be
  **much** better than with stock 2.2.x, at least
  with hardware similar to mine...

-----------------------------
But now back to the topic...

Andre: What about 2.4.x? Will it contain your ide-patch or some subset? I
hate to ask this but I don't have the time to play with 2.4-pre although
I'd like to.

If yes, forget about all this, since 2.4 will be here soon and most users
will sooner or later upgrade to it.

If not: What about starting a poll like collecting 2.2.x (and maybe 2.4.x
as well) IDE performance data with some shell script (my pleasure to
provide one)? Anything special you want to see in the data? I think about:

- /proc/version
- /proc/cpuinfo
- /proc/pci/pci
- /proc/ide/*
- Output of 'hdparm -Tt' on all IDE drives.

With the data maybe we see your ide-path in a future 2.2.x kernel or at
least some subset for onboard IDE chipsets? And I think it **HAS** to be in
2.4.x once this kernel is stable...

Thanks.

P.S.: Please email me directly, I'm not subscribed to any Linux list.

Mit freundlichen Gruessen / best regards

Michael Kwasigroch
FaxPlus/Open Development
________________________________________

eMail: mkwasigr@intercope.com

INTERCOPE GmbH

                                                                                                                                             
                    Mark Hahn
                    <hahn@coffee.psychology.mc To: Michael Kwasigroch <mkwasigr@intercope.com>
                    master.ca> cc:
                                                      Subject: Re: IDE disk slow? There's help...
                    21.10.00 04:47
                                                                                                                                             
                                                                                                                                             

> old triton 2 board (Intel 430HX) and I didn't want to risk more trouble

basically ancient hardware.

> Linux (stock 2.2.17) could ony push about 2.6 MB/s "through" it (hdparm
-Tt
> /dev/hda)... :-(

obviosly in some primitive PIO mode.

> The scsi disks can do about 5.5 - 6.1 MB/s (8Bit fast SCSI, no ultra,
> adaptec 2940 PCI).

it's not hard to come close to the 10 MB/s limit for even this many-year
obsolete scsi mode.

> My new IDE disk now "flies" at about 9.2 MB/s and really outperforms the
> scsi disks!!!

modern ide disks all sustain ~20-36 MB/s. most modern scsi disks are
about the same, with 10K and 15K a little faster.

> One thing I don't understand: Why is this patch not in the stock kernel?
It

you're using an obsolete kernel and in obsolete kernels, you need to
explicitly enable DMA modes.

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/



This archive was generated by hypermail 2b29 : Mon Oct 23 2000 - 21:00:20 EST