[linux-lvm] Apparent performance degradation for each PV with striping

[Russell Coker]

On Monday 19 March 2001 14:44, Steven Lembark wrote:
> Escalade is a nice idea, but still runs afoul of the interrupt
> & bus design of PC's.  for high-i/o applications "PC" hardware
> doesn't work all that well.  main problem here is delivering the
> raw datat to an Esc. controller through the existing PC.  the
> add'l cpu load in this case comes from the bottom-half of drivers
> on heavily loaded cards that spend too much time waiting for
> access to the shared card.
>
> these are a distinct improvement over stock IDE or software RAID
> but don't expect them to suddenly turn your PC into a SparcServer
> or K400.

What is a K400?

Every test that I run shows SPARC machines running slower than desktop 
machines with IA32 CPUs.  My Athlon-800 machine has more memory bandwidth 
than an E450 according to the industry standard "streams" benchmark and 
according to a little memory benchmark I wrote.  Also a single IBM 46G ATA 
drive in my Athlon outperforms A1000 arrays in E450 class hardware.

> note on the benchmark:  you can push the CPU to 100% w/ the dd
> of=/dev/null

What type of hard drive and what type of CPU?

In terms of using CPU power through IO my record was >80% of the power of my 
Athlon800 when doing "cat /dev/zero > file", then I fixed cat to use 4K 
buffers instead of 1K and it used significantly less CPU power.

I suspect that a large part of this was due to inefficiencies in ReiserFS in 
this regard.  It's still on my to-do list to do some more comprehensive tests 
of this issue with other file systems and with raw devices.

-- 
http://www.coker.com.au/bonnie++/     Bonnie++ hard drive benchmark
http://www.coker.com.au/postal/       Postal SMTP/POP benchmark
http://www.coker.com.au/projects.html Projects I am working on
http://www.coker.com.au/~russell/     My home page