Monday, June 26, 2006

HSM on NetWare part 2

Over at Novell's Cool Blogs Richard Jones has posted a pair of entries related to HSM. I covered the first one last week. This appears to be a series on HSM in general. Richard Jones was the presenter for IO 202: Data Protection Concepts at BrainShare this year, which I attended.

It was in IO 202 that concept of 'Shadow Volumes' was discussed. Shadow Volumes, as they were explained, combines the best features of Distributed File Systems and HSM. He didn't go into very much detail, but in essence I believe Shadow Volumes work by migrating old files to different file systems and use DFS features to allow instant access to that data without all the fun migrate/demigrate problems. Secondly, Shadow Volumes work on last-modified-date rather than last-accessed date; last-accessed date being an increasingly unreliable data-point in this era of Google Desktop, Spotlight, Beagle, and Vista Search.

To give you an idea as to how scanty my information is, these are my notes from the session, as transcribed directly from my note-book:
- Shadow Volume (not shipping) combines DFS & HSM to a degree. (Would be good)
Not a lot. Going to the sessions slides now that I have them, I find more data! And this key phrase: "Like DFS, but at a file-level instead of a Directory level." Ahah! So you take two different file systems, and some DFS-like technology unifies the directory listing of both to present a complete file-system. Then the DFS-like technology in the client is smart enough to be told which actual file-system the file lives on and go retrieve it from there silently. That way you get around the migrate/demigrate lag that traditional HSM involves, and the associated disk-space hit. These days hard-drives are cheaper than tapes on a per-gig basis for certain technologies, as the backup software vendors will tell you at great length, so it makes sense to migrate from your Fibre-Channel SAN to a SATA JBOD for instance.

What's also important about this file-level DFS is that it permits different backup policies on the two physical file-systems. The old/slow file-system can be backed up once a month, where the new/fast file-system gets some attention every day. "What's the big deal," you ask, "that's what the 'Archive' bit is for. Software has been doing that for years. Decades." By splitting your data this way, you can backup a whole volume in complete, rather than a whole volume backup that requires the software to scan the meta data of every single file and directory to check to see if that data needs backing up. That will speed up backups.

Shadow Volumes will probably require a change on the Client to make work. So we're still probably a few years away from a release involving the technology. Though, it could also work if the server itself handles the redirection... hmm. We'll have to see how it pans out.

Tags: ,

Labels: ,


Comments: Post a Comment

<< Home

This page is powered by Blogger. Isn't yours?