> -----Original Message-----
> From: Testbed Support for GridPP member institutes [mailto:TB-
> [log in to unmask]] On Behalf Of Ben Waugh
>
> I have only used Lustre as a user, not installed it myself, but unless it
> has changed drastically in the past year or two, it is not really suitable
> for a general-purpose filesystem. It is optimised for parallel access to
> large data files, and is REALLY sluggish if you are accessing larger
> numbers of small files, e.g. compiling and linking large applications or
> even listing directories.
>
Oxford has a Lustre filesystem as a fairly general purpose FS in our
local non-grid cluster, and it does pretty well. We did have some
issues in the early days with an older lustre release and a really
underspecced metadata server, but the current generation is fine.
It's notionally our ATLAS group's 'data' disk, but they put all sorts
of stuff in there - it's far from being just well organised bulk files.
Indeed a quick divide of the space used by the inodes used shows the
average file size to be about five megabytes, which, given that there
are some pretty big data files as well, gives you some idea that there's
a lot of small stuff in there too dragging the average down.
Potentially interestingly, we do have both the Lustre system and some
NFS space built out of very similar hardware; it might be fun to try
(say) some big compilations on each of them and see how it goes.
At this point I really would say that Lustre is the sensible choice in
this area; I wouldn't be surprised if one of the shiny new upstarts
takes over the front runner's position in a few years, but they haven't
yet.
Ewan
|