[ILUG] Dell PERCs and Linux

[David Murphy]

Quoting <Pine.LNX.4.21.0006160019540.5791-100000 at fogarty.jakma.org>
by Paul Jakma <paul at clubi.ie>:

> let's see. each chunk won't be more than 128kb/64kb, and the drive
> will have at least 512kb of cache. so that can be go straight into
> cache, and the drive can disconnect, controller can feed another
> drive, the original drive can post the write later. 

The drive buffers are volatile RAM, therefore the application can't be
told the write has suceeded until it's been written to the physical
disk.

> > 2) If you have three SCSI controllers, one disk on each, then you
> > still have to copy the data to two controllers instead of three.

> same for RAID1...

Gah, that should have been 3 controllers for 3-column RAID5 with one
disk per controller, versus 2 controllers for 2-column RAID1

> > 3) No matter what, you have to calculate the RAID5 parity info
> > before you write it.

> ok.. i am assuming that the RAID card's CPU is fast enough, that
> this is not a factor. Which is true for most newer cards, (eg
> ExtremeRAID has a StrongARM @ 100MHz+) and for linux software RAID.

No matter how fast it is, it can't do it instantaneously, so it will
*always* be slower than something that doesn't have to do it.

> > > no, the 3disk RAID5 has double the read throughput of a 2 disk
> > > RAID1.

> > How do you work that one out?

> by observation. bonnie doesn't lie.

Perhaps your RAID1 implementation is lying, or rather doesn't
round-robin reads from both sides of the mirror?

-- 
When asked if it is true that he uses his wheelchair as a weapon he will reply:
"That's a malicious rumour. I'll run over anyone who repeats it."
Stephen Hawking - [http://www.smh.com.au/news/0001/07/features/features1.html]
David Murphy - For PGP public key, send mail with Subject: send-pgp-key