> If you consider creating a huge filesystem.... Eg stripe or concat all
> your LUNs together, and put all your DBO devices on this filesystem as
> subdirectories.... It does work, but with UFS you would be in serious
> trouble if you had to do a fsck....and of course you would use UFS
> logging, and that would concentrate all the DBO device's write latency
> to the region which is handling logging...
UFS is only logging metadata, not data. It should be almost
insignificant if you're only accessing a few files.
> So... UFS is not cool, either way. And vxfs is just as bad, or worse....
> What I am looking at now is Sun's SAM-FS.. With it you create a device
> which is made out of a pool of LUNs for storing data, and another pool
> for storing metadata...
>
> So you can create some small high performance devices... Using RAID0+1
> (striped/mirrored) (or RAID1+0 depending on religion).... Use these for
> metadata (and logging I believe)... Create the rest as RAID5 or RAID3
> and assign as data devices, SAM-FS manages how the data is spread
> between the data devices, and is configurable... I am still trying to
> see if you can 'ice' a device temporarily eg when you've lost a disk and
> you want to minimise writing while reconstruction takes place...
>
> Mount this device up as one enormous filesystem (or one per Pool)....
> Create subdirectories for each DBA device... How many adv_file devices
> you have depends on how many tape drives you have and your networker
> license, but I would recommend having one more tape drive device than
> adv_file devices so that you can stage to tape from all adv_files at
> once but still have one drive free for restores....
>
> Using SAM-FS like this is not too expensive.... The sun licence for
> SAM-FS is per server (note that you don't want to pay for the SAM-QFS
> licenses which allow shared access and archiving and cost a LOT of money
> and are based on number of terabytes).
ZFS should have similar benefits, but is not yet available in a
production release (but will be free when it is). Metadata is not
concentrated in one area of the filesystem, but is spread throughout.
--
Darren Dunham ddunham AT taos DOT com
Senior Technical Consultant TAOS http://www.taos.com/
Got some Dr Pepper? San Francisco, CA bay area
< This line left intentionally blank to confuse you. >
To sign off this list, send email to listserv AT listserv.temple DOT edu and
type "signoff networker" in the
body of the email. Please write to networker-request AT listserv.temple DOT edu
if you have any problems
wit this list. You can access the archives at
http://listserv.temple.edu/archives/networker.html or
via RSS at http://listserv.temple.edu/cgi-bin/wa?RSS&L=NETWORKER
|