On Mon, 27 May 2002 18:57, Robert Sander wrote:
> Hi!
>
> We will have a new RAID-System shipped the next weeks with a total
> cpacity of 1.6 TB. I want to run XFS (what else?) on it and have some
> config qeustions.
> The RAID consists of 12 160GB IDE-Disks connected via SCSI to the host
> computer. The host only sees one large SCSI disk
Therefore, I expect that all RAID is done on the array hardware.
> I think that making just one partition on that disk and running a
> mkfs.xfs without any special options will not produce optimal
> performance.
This depends on many factors - number of files, size of files and access
pattern on files, ratio between writes and reads etc...
I don't have experience running filesystems that big - some of the gurus on
the list have mentioned increasing the log size a little, as well as setting
swidth during filesystem creation. The gurus on this list will be able to
give you better info than this.
> Should I create a separate logfile partition on the RAID? How large
> would it be? What are other options for mkfs.xfs that I should look
> into?
An external log should not be needed AFAIK. The performance problem with
RAID5 & XFS only exists when the RAID is being done by the Linux MD software
RAID5 code. XFS log writes do not interact well with the RAID5 code and the
RAID5 code flushes buffers all the time - therefore, the performance hit.
(An indepth analysis you'll be able to find on the mailing list somewhere -
last year if my memory is correct).
>
> Thanks for the answers. Any pointers to the FAQ are also appreciated.
>
> Greetings
Dan Yocum has a webpage somewhere listing all sorts of goodies in relation to
large filesystems.
http://home.fnal.gov/~yocum/storageServerTechnicalNote.html
There are others but my mind has drawn a blank at the moment. (0200 here at
the moment). Other people might be able to help you out.
--
Adrian Head
(Public Key available on request.)
|