> > > I would prefer to play with the AAA-133
> > > three port card.
> >
> > 3950U2B and a 64bit PCI bus would be much nicer :)
>
> True but pragmatically, I have no machine with 64 bit PCI and I have
> nine drives in a stripe and 3 SCSI channels (3 drives to a channel)
> has been a useful confuguration excepting that it wastes two PCI slots,
> so I would like to try the AAA-133, is it supportred?
The AAA133 is not supported. We don't do anything with those on board RAID
chips on those controllers. I don't know if we end up finding the channels
behind that RAID chip or not, but I suspect we don't. As for 64bit PCI, you
don't have to have a 64bit slot to use the 3950U2B (I don't) but if you have
one, the card is fully 64bit capable. It's also dual channel, so it's
better than one channel per slot.
> > Depending on the kernel version and such, I can get some pretty nice
> > performance figures:
> > [root@dledford /mnt]# bonnie -s 256
> > File './Bonnie.2021', size: 268435456
> > Writing with putc()...done
> > Rewriting...done
> > Writing intelligently...done
> > Reading with getc()...done
> > Reading intelligently...done
> > Seeker 1...Seeker 3...Seeker 2...start 'em...done...done...done...
> > -------Sequential Output-------- ---Sequential Input--
> > --Random--
> > -Per Char- --Block--- -Rewrite-- -Per Char- --Block---
> > --Seeks---
> > Machine MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec
> > %CPU
> > 256 11370 98.9 52781 95.8 16791 83.3 12767 89.5 32351 78.4 513.0
> > 14.0
>
> I trust that the memory size of the machine is much less than 256 MB,
> if so, this is quite nice. How many and what kind of disks?
The memory size is 128MB. This was a quick test to demonstrate, but I get
roughly the same results with a 1024MB file size (reduce the seeks/sec
number since the larger file gives it more space to seek and slows that
particular test down). It's on a dual PII266 with 5 Seagate Cheetah Ultra2
Wide drives on one 3950U2B controller. This is a somewhat less than optimal
setup. Given the right combination of chunk sizes, ext2fs block sizes, and
read-ahead I was able to get those figures up to 60MB/sec output and
42MB/sec input. I'll see if I can't find those numbers again....hang
on.....<pause as I remake the raid0 array and filesys to try and find that
combo again>....OK....using a chunk size of 32, 5 disks, 4k ext2 block size,
stride=40 and a filesize of 1024MB, I get this:
-------Sequential Output-------- ---Sequential Input-- --Random--
-Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks---
MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec %CPU
1024 11827 99.5 54637 96.9 18327 90.5 13498 94.0 39208 93.8 369.9 10.5
I used to be able to get slightly higher on both the read and write, but the
difference now could be either the fact that I'm farther back on the disks
now due to partition changes, or it could be that I've since added more
checks into the aic7xxx driver that will get disabled when I'm in production
mode and I've upgraded to Ingo's latest raid code. Anyway, I'll be
re-running the tests when I've streamlined the aic7xxx driver to see what
happens.
--Doug Ledford <dledford@dialnet.net> Opinions expressed are my own, but they should be everybody's.
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/