On Thu, 18 Apr 2002, Wienand Ian wrote:

> firstly what file system are you using?  i can't see this working with any
> standard non distributed file system.

It will work with most filesystems as long as the two machines coordinate
meta-data usage.  It may not be quick with some filesystems.

> i've never seen anything like this -- two pc's essentially using a single
> disk.

DEC clusters have had this feature for years, which is exactly why SCSI 
allows it.

> i can't see how this setup can guarantee normal unix semantics

You need to run a 'distributed lock manager' to share and sequence
meta-data changes and to retain file lock semantics.  There are a few
projects working on this, and Compaq recently contributed DEC's DLM code
to the high availability Linux effort.

>  the scsi bus handles putting blocks onto the disk, and that is all

SCSI also allows the disk itself to run a filesystem.  See the Object
Based Storage Devices Command Set.  Of course, your disk has to support
OSD for this to be an option, and only disks in supercomputer clusters do
this today.  But now the code has been written it's only a matter of time
before it appears on commodity storage.

The choices today are to be cutting-edge and run a DLM.  Or to run a
networked file system on one of the machines (with a necessary performance
hit).  Or not to use both disks hosts at once (eg: to allow failover
rather than load-sharing).

-- 
Glen Turner                                 Network Engineer
 (08) 8303 3936      Australian Academic and Research Network
 [EMAIL PROTECTED]          http://www.aarnet.edu.au/
--
 The revolution will not be televised, it will be digitised



-- 
SLUG - Sydney Linux User Group Mailing List - http://slug.org.au/
More Info: http://lists.slug.org.au/listinfo/slug

Reply via email to