> Here is a potential performance problem uncovered
> by a set of benchmark tests.
> Namely two SCSI adaptors don't speed up raid 0/1 perceptibly
> in comparison to the performance of a single SCSI adaptor system.
> (and under dual CPU system for that matter .)
Hi, do you know how many disks he was spinning? It looks like just two
drives, in which case, there will NO real advantage on having dual SCSI
controllers, since the SCSI bus is NOT the bottleneck. The primary reason
for dual scsi controllers in this instance is to insure that any failure
that completely kills one scsi buss doesn't kill the other. I.e. cabling,
fried SCSI card, a drive that holds a line at ground forever and ever amen.)
Assuming the SCSI controllers are using UW or U2W busses, the max throughput
on the buss should be 40 or 80 MBytes/second. A single server class hard
drive will likely have a throughput of between 5 and 15 Megs a second.
Judging by the numbers listed in the RAID-0 tests would appear to be from a
pair of hard drives each pushing out about 6-7 Megs a second.
Further, since Ultra SCSI is usually limited to a maximum of 1.5m cable run
length with about 0.25m to 0.30m spacing for stability reasons, the maximum
number of devices, counting the controller tops out at about 7 devices (6
hard drives), your hard drives would have to pump out 11+ Megs/sec to
saturate a U2W card or half that for the UW cards.
So, on a U2W, until you fill one card with 6 drives, there's probably no
performance advantage til you add the second card. Then you should see a
difference. So you probably need to spin 10 or 12 fast hard drives
on multiple controllers to really see a gain from the second card. The gain
will likely NOT be linear, but each drive should speed things up a large
percentage of it's raw speed in a RAID 0.
My testing on Linux RAID-1 is that it is no faster reading right now than
the single drive, just more reliable.
-
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/