| T.R | Title | User | Personal
Name | Date | Lines |
|---|
| 261.1 | | COOKIE::FROEHLIN | Let's RAID the Internet! | Thu Mar 28 1996 15:37 | 25 |
| 261.2 | Thanks for your quick reply ! | ORO50::REEVES | Fire and Forget. | Fri Mar 29 1996 06:52 | 9 |
| 261.3 | | AMCFAC::RABAHY | dtn 471-5160, outside 1-810-347-5160 | Tue Feb 25 1997 07:46 | 10 |
| IDX's ZBACKUP is a source of pain. What would it take to make;
$ MOUNT /SYSTEM DSA1 /SHADOW = DPA137: TEST
work? I gather the shadowing code is the one rejecting the DP device. Is there
some fundamental reason for it to do this? Is there some bit that can be set to
fool the shadow code into accepting the DP device?
Yes, RAID 1+0 is vastly inferior as compared to RAID 0+1 but until IDX enhances
their ZBACKUP to make use of the RAID CLONE command we're stuck.
|
| 261.4 | | COOKIE::FROEHLIN | Let's RAID the Internet! | Tue Feb 25 1997 09:13 | 8 |
| What is ZBACKUP?
Shadowing code refuses to work with a DPA device. Haven't looked into
details.
What's the discussion they have about RAID 0+1 or 1+0?
Guenther
|
| 261.5 | | AMCFAC::RABAHY | dtn 471-5160, outside 1-810-347-5160 | Tue Feb 25 1997 10:08 | 20 |
| ZBACKUP is a DCL command procedure supplied by IDX. It hides the details of
invoking the VMS BACKUP utility. It can freeze the application, flush
everything, split shadow sets and unfreeze the application. It launches
concurrent BACKUP's and then reforms the shadow sets as they finish.
ZBACKUP was developed years ago before RAID anything and has grown
incrementally. IDX loves RAID-0 as it spreads the load more evenly than volume
sets which is how they use to support large data sets. Controller-based RAID-0
is the only thing they've "integrated" with. Host-based RAID-0 is not workable
for them as it stands.
RAID-1 is extremely common in IDX configurations. It is the only way they can
do an "online" backup. Their application is well integrated with hosted-based
RAID-1. There is a configuration file which maps DS devices to physical devices
so that they can know which to dismount and remount. It also directs which
disks backup to which tapes.
So, they end up with RAID 1+0. Controller-based 0 and host-based 1. This isn't
so bad with modern disks having such huge MTBF's. Still, RAID 0+1 would be
better but they would have to do the work in their ZBACKUP to support it.
|
| 261.6 | | COOKIE::FROEHLIN | Let's RAID the Internet! | Tue Feb 25 1997 11:17 | 6 |
| Doesn't look like lots of changes to me since the RAID CLONE command
supports the /USE_DEVICES=(unit-list) which allows to specify which
shadow set members to break out of the original shadow set. But needs
some thorough testing, though.
Guenther
|