[GE users] File server overloading

tmac tmacmd at gmail.com
Wed Jan 30 20:35:34 GMT 2008


    [ The following text is in the "ISO-8859-1" character set. ]
    [ Your display is set for the "ISO-8859-10" character set.  ]
    [ Some special characters may be displayed incorrectly. ]

I have 200 grid nodes.
Each node is connected with GigE to a 4948-10G
Each 4948-10G is uplinked to a Cisco 6509 via 10Gige
I have a 2-head NetApp GX system connected with 2 x 10GigE

I get 2-3 Gigabytes/sec from my netapp to the cluster.

ps. I have tuned sysctl parameters and tweaked the NFS mount options
All systems use the same storage.

Very easy and scalable storage.

--tmac

On Jan 29, 2008 9:29 AM, Rayson Ho <rayrayson at gmail.com> wrote:

> I agree with Andre. And if you are already using GFS, then you should
> find Lustre familiar.
>
> In fact, I saw this white paper a week ago:
> http://www.synopsys.com/products/solutions/dfm/hpc_cats_wp.pdf
>
> Lustre is used by TOP500 supercomputers, and also other industries...
> I just updated the list last night:
> http://en.wikipedia.org/wiki/Sun_Microsystems#HPC_solutions
>
> And as Lustre is now owned by Sun, I believe Lustre would grow in a
> faster rate than other shared disk filesystems -- Grid Engine is just
> a nice example of how much effort Sun put into an opensource project!!
>
> (Off Topic: when Sun completes the MySQL deal, I believe MySQL would
> also grow in a much faster rate too!!)
>
> Rayson
>
>
>
> On Jan 28, 2008 7:18 PM, Andre Gauthier <gauthier at cs.umass.edu> wrote:
> > I suggest you take a look at lustre.  We had good luck with it.  It is
> > faster than any shared filesystem that I tested and scales really well.
> >
> >
> > Gary Thomas wrote:
> > >
> > > Does anyone have experience with using Redhat's GFS with SGE.  We are
> > > having problems where
> > > 15-20 execution hosts running jobs that reside on the same file server
> > > are overloading the server.
> > > We have 3 file servers and are thinking about using GFS to combine
> > > them into a file system so
> > > We can have all 60 of our execution hosts running off the same virtual
> > > file system.  We're looing
> > > At expanding to over 100 execution hosts in the near future and need a
> > > file server system that can
> > > Keep up.
> > >
> > > Any suggestions would be appreciated.
> > >
> > > GT
> > >
> > >
> > >
> > > Gary Thomas
> > > Senior Software Engineer
> > > (425) 869-4227 x716
> > > gthomas at forteds.com
> > >
> > >
> > >
> >
> > ---------------------------------------------------------------------
> > To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
> > For additional commands, e-mail: users-help at gridengine.sunsource.net
> >
> >
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
> For additional commands, e-mail: users-help at gridengine.sunsource.net
>
>


-- 
--tmac

RedHat Certified Engineer #804006984323821 (RHEL4)
RedHat Certified Engineer #805007643429572 (RHEL5)

Principal Consultant, RABA Technologies



More information about the gridengine-users mailing list