Re: distributed shared memory / mmap

[Louis Rilling]

On Wed, Aug 20, 2008 at 03:43:24PM +0100, Steven Whitehouse wrote:
> Hi,
> 
> On Wed, 2008-08-20 at 15:18 +0200, Martin Uecker wrote:
> > 
> > Hi!
> > 
> > Is there any easy way to get shared memory between to
> > processes on different machines? It seems possible using mmap
> > on a file system like GFS on top of a distributed block device.
> > 
> > 
> > Sincerly,
> > Martin Uecker
> > 
> Yes, you can do it using GFS/GFS2 and it should work. On the other hand,
> don't do it. Its probably not what you want and unlikely to give you
> good performance since the locking is "per inode" and thus you'll
> generate a lot of cache flushing if there is any appreciable amount of
> writing to the shared region. If you have a read mostly (or only)
> workload, then it will work much better.

Alternatively, you can use kernel-level DSM like provided by Kerrighed.
Kerrighed provides cluster-wide SYSV shm segments, which sharing is handled
at the page level -> far better than the inode level.

You may find this solution not-so-easy though...

Regards,

Louis

-- 
Dr Louis Rilling			Kerlabs
Skype: louis.rilling			Batiment Germanium
Phone: (+33|0) 6 80 89 08 23		80 avenue des Buttes de Coesmes
http://www.kerlabs.com/			35700 Rennes

Attachment: signature.asc
Description: Digital signature