Re: Interpreting disk-zone-profile

From: Jens Axboe <>
Date: Fri, 20 Jul 2007 14:12:03 +0200

On Fri, Jul 20 2007, Erwan Velu wrote:
> Jens Axboe wrote:
>> [...]
>> Are you sure you used the new base?
> I thougth I was and I did a double checking... And I wasn't on the new
> base...
> What a shame on me...
> As you can imagine... your patch solved this issue.
> Thanks for this quick patch and sorry for the false positive report ;)
> So now, the performance is falling down slowly (sounds good) but I my case
> the job stopped at 90%.

ETA is just a guideline, it cannot be really precise always. So not much
I'll do about that.

> So the 10% remaining represents 15G meaning there is enough place to do
> some other tests as we skip 2G at each turn.
> Any idea why it stops at this size ?

Did you check whether it didn't look at the lsat 15g, or is that just a
guess based on the 10% number you quote above?

> I also wonder how "per" can reach 100.12% as you explain it shows the speed
> threads can take from the same disk. Here I'm running on the same disk with
> 1 thread so I can't expect reaching more than 100 % right ?

Rounding errors :-)

Jens Axboe
Received on Fri Jul 20 2007 - 14:12:03 CEST

This archive was generated by hypermail 2.2.0 : Fri Jul 20 2007 - 14:30:01 CEST