> I was under the impression that z/VM 5.2 now allowed native use of SCSI > DASD, which I assummed also ment SCSI TAPE.
Nope -- that's why I'm pushing this issue. zVM 5.2 supports only FBA (9336) emulation (which limits use of SCSI disk to VM, Linux and VSE), and no SCSI tape support at all for CMS or CP use. SCSI tape is supported only for direct Linux guest attachment, and TSM and DDR don't support backing up CMS minidisks to Linux-attached tape. Neither does VM:Backup or HIDRO. > In my case, I'm in the process of cost justifying an IBM VTS/3494 > setup. That is virtual enough for our requirements (high capacity, fast > transfer and no Operations). > I've been wanting virtual tape for over 5 years now. That is a decent > virtual tape (the MP3000 emulated tape just wasn't so good for writing). > And VSE Virtual tape over the network (10 mbs network speeds), didn't > pan out either (didn't help that Dynam didn't really support them well). > But we have a Linux project that has the added justification of a tape > upgrade. Just can't backup 4 TB with our bus and tag IBM 3490A/B > drives. You are aware that you can't use a channel-attached 3494 or VTS from Linux w/o getting some other OS to handle the tape mounts, and that IBM does not currently supply ANY support for the library devices (see our tape mount daemon for a way around that) in Linux? And that if you SCSI-attach some drives in the 3494, you can't use them from the non-SCSI aware OSes (the controller boards on the drives are physically different)? Partitioning ATLs is a royal PITA both technically and operationally. > But back on the origional topic, I would gather that larger shops > already have solved their tape problems with a robotic setup (perhaps > with VTS). The small shops (P/390 and Flex/ES shops), already have > their solutions. It is the mid size shops that may be able to > influnance the use of the resources to have emulated tape on disk. It's actually surprising how many of the larger shops are facing power or space restrictions. Adding a new ATL just because the Z can't deal with the one they already have is a show-stopper in many places...wouldn't it be an easier case to make if the new ATL you're talking about buying could be used directly by both Z and open systems w/o reconfiguration? Or you could use the big SDLT or LTO libraries you already own on the open systems side? As you say, the flex folks have a solution -- emulated tape. The large folks actually don't have a good answer (or don't care), and the midsize folks are (as usual) stuck in the middle. -- db
