[prev in list] [next in list] [prev in thread] [next in thread] 

List:       netbsd-users
Subject:    Re: vm.bufcache and git
From:       Greg Troxel <gdt () ir ! bbn ! com>
Date:       2010-05-27 12:34:59
Message-ID: rmi6329plos.fsf () fnord ! ir ! bbn ! com
[Download RAW message or body]

Content-Transfer-Encoding: quoted-printable

Adam Hamsik <haaaad@gmail.com> writes:

> On May,Wednesday 26 2010, at 7:42 PM, Greg Troxel wrote:
>
>> On a system with big enough bufcache and maxvnodes, I found stat to take
>> about 4 us up to about 32 files, and then it starts to grow, reaching
>> 9-10 us around 4096-8192 files, and then it remains relatively constant.
>>=20
>
> How many times have you repeated this test. Maybe there was some
> external influence (cron, other program etc.. ) who mixed with your
> results.

I ran it several times on several systems and the pattern was
consistent.  There is definitely a fair amount of measurement noise, but
I think the basic take-away that stat over many files takes longer is
accurate.  It's also not surprising, as more files causes the data to no
longer fit in caches.

I would be curious to see results on other systems.


[Attachment #3 (application/pgp-signature)]

[prev in list] [next in list] [prev in thread] [next in thread]