"David" == David H Brierley dhb@ssd.ray.com writes:
David> On Wed, 12 Apr 2000, John Stoffel wrote:
- Are you willing to spend the money on NDMP capable software and a
DLT7000 jukebox attached directory to the filer(s)? That would get the network out of the equation almost completely.
David> I've seen several comments like this in the past and I don't David> understand them. Why do you feel it is neccessary to spend David> money on NDMP capable software to be able to use a directly David> attached tape drive?
Because I want to be able to do all my backups to one jukebox if at all possible. This keeps down costs, keeps down maintenance issues in terms of where you have to go to swap tapes, etc. In our situation, I have filers on four seperate server closets (they're not rooms, dammit! :-) so I don't want to have to walk around the building to swap tapes.
Also, I can see that I wasn't quite clear in my previous statement, since it will get the NetApp's backup traffic off the network, but if you use the NetApp as the NDMP target for other hosts, their traffic will be going over the network.
And of course there's the issue of indexes, since it does no good to backup your files if you can't restore them in a timely manner, and having to dig through tapes to find a file is slow. That's why all the major backup software vendors have online file indexes of files backed up.
David> It probably took me a couple of hours to put togethor a shell David> script which will back up the filer using the locally attached David> tape drive. The script runs on the admin host and uses rsh David> commands to connect to the filer and run the dump program. Two David> hours of my time is a lot less expensive than any of the David> commercial backup solutions, especially since it would have David> taken me almost 2 hours to fill out the paperwork for the David> purchase. Plus, it was an interesting problem to solve so I David> enjoyed doing it.
It's a tradeoff. And everyone is in a different boat. I've been in both the Educational environment where we didn't have the money for backup software, so I wrote the scripts to manage backups and tapes and all. It's not trivial to do, especially since the core of it's really simple. It's the error checking and making the script resistent to errors (both user and system) that's hard. I've heard it said that 90% of all code is bounds and error checking and I can believe it. It's a hard problem to solve well.
Some of the tradeoffs include:
- how many systems you need to backup. It goes from 1 to 100 to N. The more you have to backup, the better a central tape library looks.
- Is your time cheaper than software costs? Usually this is true in the non-profit space. For those cases, I also look at the Amanda backup package.
- How much data do you have to backup? And how long can you take to back it up? Time is money in the business world.
- How fast do you need to restore a single file from backups. This is a big hidden gotcha that you need to think about. And sometimes you can't justify it until after you've been bitten.
And I'm sure people can chime in with more.
For us, the software costs alot less than my time, since I'd have to baby-sit homegrown scripts all the time to get them working, and no homegrown script offers the speed of a Legato or Veritas or other costly backup solution, they just don't do the data interleaving. Yet.
John John Stoffel - Senior Unix Systems Administrator - Lucent Technologies stoffel@lucent.com - http://www.lucent.com - 978-952-7548 john.stoffel@ascend.com - http://www.ascend.com