Hi List-Readers!
The kernel of current Linux-Distributions does not support
muliple pathes to a dasd device at all.
A workaround is to spread the data over multiple devices
using LVM or MD in striping mode. Using the same amount
of devices like the amount of pathes available (or a multiple
of it) should fit best.
This problem is already addressed in the current (experimental)
2.4.17 code.
mit freundlichem Gru� / with kind regards
Carsten Otte
IBM Deutschland Entwicklung GmbH
Linux for eServer development - device driver team
Phone: +49/07031/16-4076
IBM internal phone: *120-4076
--
We are Linux.
Resistance indicates that you're missing the point!
|---------+------------------------------->
| | Michael |
| | MacIsaac/Poughkeepsi|
| | e/IBM@IBMUS |
| | Sent by: Linux on |
| | 390 Port |
| | <[EMAIL PROTECTED]|
| | .EDU> |
| | |
| | |
| | 05/16/02 05:42 PM |
| | Please respond to |
| | Linux on 390 Port |
| | |
|---------+------------------------------->
>--------------------------------------------------------------------------------------------------------------|
|
|
| To: [EMAIL PROTECTED]
|
| cc:
|
| Subject: Re: Multipath I/O on 390 Linux
|
|
|
|
|
>--------------------------------------------------------------------------------------------------------------|
> the OS doesn't need to know about the multiple paths
For high availability, yes. But for performance, I was *under the
impression* that Linux needs to be fooled into using the multiple
paths (haven't been able to confirm this with end-to-end performance
tests). This is done by LVM or raid-tools striping (RAID 0). *As I
understand it* the "fooling" works as follows - when a striped
volume is detected, the Linux kernel will continue with data
transfers before the previous one finishes. Then the multiple
I/O paths to the DASD will be utilized. Actually the first time
I tried a performance test, I saw a small performance gain, but it
was negligible enough to be noise.
The recently I noticed in make menuconfig the setting:
Multi-device support (RAID and LVM) --->
<M> Multipath I/O support
which is not always on. So I'm hopeful for some serious performance
gains using RAID 0 and a kernel with this setting on. Any comments
from performance guys with a better background on this?
-Mike MacIsaac, IBM [EMAIL PROTECTED] (845) 433-7061
