Thanks for the tip, JAM. It sounds like good news!
On 04/20/16 13:31, John McKown wrote:
What really seems a bit strange is that it is touted as a way to share
data with a "distributed" system, but doesn't say how. E.g. There is a
Windows server which has a NTFS file system on a LUN. That LUN can be
accessed via zDDB for z/OS. OK. This would mean that the z/OS program using
this interface would need to implement the NTFS file system architecture in
its own code. The same with ext3, or ext4, or btrfs, or ??? with Linux. Or
that the LUN would not have a file system on it at all. It would just be a
"raw" device to z/OS and ??? on the SAN side and they would need to
implement their own file system time code.
* We share DASD in mainframe land all the time.
* We share "minidisks" in VM land all the time.
* It's been proven that virtual Linux (any platform, not just z)
can share disks (LV, disk "images", or physical devices).
* Evidence suggests that disk sharing can be extended to other systems.
(Maybe even Windoze, but certainly Unix systems can cope: AIX,
Solaris, HP-UX, the BSDs.)
SAN is to the rest of the world with FICON DASD is to the mainframe.
Wow ... what a concept!
Let the sharing begin.
* EXT2 can be shared R/O among systems (virtual, real).
* EXT3, 4, and BTRFS are journalling and tend to attempt replay even
when mounted R/O.
(Force the underlying volume R/O and ignore the replay errors.)
* dunno NTFS all that well, but isn't it journalling? good luck
* ISO9660 can be shared among systems (virtual, real, even unlike op
sys!). The FS is R/O by nature.
* A LUN with no FS at all is "just a file" to any Unix. Read it. Write
it!
R/O sharing is safe.
Single R/W with multiple R/O can be done with care. (Think LUN switching
or minidisk flipping.)
R/W filesystems can be shared when some form of media locking is
available. To date (as far as I know) such locking mechanisms have been
out-of-band. (See OCFS2 and GFS2 for examples. Please correct me if I'm
wrong, if either of them can do in-band locking.)
If I seem a little excited about the idea, well ... you betcha!
See XIPFS mod to EXT2FS for memory based sharable block storage. (But
try that with NAND flash? Good luck with that.) And y'all thought
hipersockets were fast.
-- R; <><
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN
