Re: FreeBSD 5.3b7and poor ata performance

On Oct 25, 2004, at 4:45 PM, Scott Long wrote:
>> Someone who uses RAID-0 or RAID-1,0 modes really does expect to see a 
>> performance improvement.
>
> RAID-0 yes, RAID-10 no, at least not for software RAID.  The machine 
> winds up having to transfer the same data twice across the PCI bus, 
> twice through the controller, etc.  If the controller is on a simple 
> PCI-32/33 bus then it will quickly become saturated.

Hmm...sure, for that set of conditions-- write access, software RAID, 
bus throughput < aggregate raw disk performance-- you're quite right.

Any RAID system which is being limited by PCI bus throughput was not 
spec'ed out particularly well, the prevalence of IDE RAID on the 
motherboard via Promise or Highpoint controllers not withstanding.

However, such a configuration can still gain significant performance 
improvement from RAID-10 in terms of splitting seeks between 4 or more 
spindles, which gives you a very high # of I/O ops per second, and 
reduces service latency-- what Solaris calls "res_t", for time an I/O 
op is resident in the queue before actually being issued to a device.

Even if you suffer from bus throughput limitations, and are not limited 
by IOPS, the read performance of RAID-10 is still very nice.

> Anyways, having spent a good part of my career with RAID, I find that 
> I only use RAID-0 when I want to test system bandwidth, not when I 
> want to store data.  YMMV =-)

I'd agree with that.  RAID-0 isn't worth the convolution of data 
compared with direct access and dividing the workload between spindles 
some other way.  However, given my druthers I'd far rather set up four 
or six drives in a RAID-10 config than in a RAID-5 config.

-- 
-Chuck