Boston Linux & Unix (BLU) Home | Calendar | Mail Lists | List Archives | Desktop SIG | Hardware Hacking SIG
Wiki | Flickr | PicasaWeb | Video | Maps & Directions | Installfests | Keysignings
Linux Cafe | Meeting Notes | Linux Links | Bling | About BLU

BLU Discuss list archive


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

block network devices



On Tue, Jun 23, 2009 at 11:46:32PM -0400, Bill Bogstad wrote:
> On Tue, Jun 23, 2009 at 10:35 PM, Dan Ritter<dsr-mzpnVDyJpH4k7aNtvndDlA at public.gmane.org> wrote:
> >
> > I've been looking at cluster filesystems over iSCSI and AoE, and
> > on moderate hardware, performance appears to be execrable.
> > Anyone around have experience with this sort of thing?
> >
> > For a simple example, I have:
> >
> > - a network that I can measure 93 MB/s across
> 
> > - a disk that I can measure 74 MB/s reading
> > - a best-case ext3 over AoE of 41 MB/s
> 
> It seems to me that you are adding both network and filesystem
> overhead here.  I just ran some quick 'dd's and I get similar
> performance drops between direct reads of /dev/sd* and reads of large
> files locally from an ext3 filesystem.  Interestingly,
> I have another partition on the same disk formated as a  XFS
> filesystem and local reads of a large file from that filesystem
> gave me around 70 MB/s.  You might try AoE with XFS rather then ext3
> and see what you get.

I can get 73MB/s reading off the ext3 filesystem, so that's not
it. In any case, I don't actually want ext3 or xfs; I'm going to
need a cluster-aware filesystem.

-dsr-

-- 
http://tao.merseine.nu/~dsr/eula.html is hereby incorporated by reference.

You can't defend freedom by getting rid of it.






BLU is a member of BostonUserGroups
BLU is a member of BostonUserGroups
We also thank MIT for the use of their facilities.

Valid HTML 4.01! Valid CSS!



Boston Linux & Unix / webmaster@blu.org