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

Re: [Linux-cluster] Recommendation of SAN devices



Unless you are working with very few very large files with low frequency
of opening/closing those files, your bottleneck isn't going to be the
SAN or the latency of the disk array, it'll be the latency between your
cluster nodes.

If you want high performance, you'll have to design your application to
play nice with performance bottlenecks of cluster file systems. Here are
a couple threads from the archives of this mailing lists to get you
started.

http://www.mail-archive.com/linux-cluster redhat com/msg04412.html
http://www.mail-archive.com/linux-cluster redhat com/msg08177.html

Gordan

On Wed, 2010-07-28 at 08:13 -0500, Peng Yu wrote:
> Hi,
> 
> Does anybody have any experience with any vendors of storage devices
> (for performance) for cluster file system? Or could you please
> recommend where I should look for relevant information? I googled, but
> there are just so much information that I don't know where to start
> with.
> 
> On Mon, Jul 26, 2010 at 2:40 PM, Peng Yu <pengyu ut gmail com> wrote:
> > Hi,
> >
> > Essentially, I want to build a cluster where multiple machines
> > (multi-core) access the same cluster file system (say using GFS). Each
> > machine might
> > concurrently access the same file system for write or read. The
> > processes running on these machines are I/O intensity. I think that if
> > each machine could use hundreds of MB/s I/O. I want the whole system
> > be scalable in the sense that
> > more machines can be added without making the I/O the bottleneck.
> >
> > The SRX model in the following webpage is a choice. However, I'd like
> > to see all the possible choices from different vendors so that I can
> > make a more informative decision. Could you please share me some
> > insights that you have or point me to a mailing list where I might get
> > some answers?
> >
> > http://www.coraid.com/products/index.html
> >
> > --
> > Regards,
> > Peng
> >
> 
> 
> 



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