[Lustre-discuss] Luns bigger than 16T

Peter Grandi pg_lus at lus.to.sabi.co.UK
Mon Jun 21 13:34:57 PDT 2010


> Hi, With 3T disks coming out soon, can anyone tell me what is
> involved with creating an OST that is bigger than 16T (on the
> 1.8.x branch)?

It is very easy: please submit patches for extending the 1.8.x
branch to use ZFS or XFS or JFS as the underlying storage layer.

> Will Lustre support 32T or 64T LUNs any time soon?

It already does, but only with one or more 8TiB/16TiB filesystem
inside them.

Part of the issue here is that large (> 4TiB) filesystem
management is largely an unsolved research problem, and while
some salesmen claim they can routinely do much larger single (as
opposed to chunked like Lustre) filesystems, they are the same
type of salemen who were selling derivatives until recently
(they are fantastic, as long as everything goes all right).

Of course as many "contributors" to this list show, there are
those persuaded that they know better, and feel lucky. Good luck
to their customers and/or employers.



More information about the lustre-discuss mailing list