BIL@su-sierra.arpa (Bil Lewis) (10/08/86)
It recently occured to me that my company spends at least $100k/year to do backups on our 8 (soon to be N+1) unix machines. [That's guess-timating human costs + tape & local storage.] We have a monthly off-site, and cycle ~100+ tapes for dailys on-site. If we purchased 1 more Eagle and did daily backups to that, we could dispense with all the local/daily stuff completely, improve response on restores (let the users do it!) and have our tape operator do more interesting tasks. Question: Is there some basic factor I'm missing, or is this brilliant? [Has someone already done this? Is it as much of a win as I figure?] thx for the thoughts -Bil (a new sys adm, but growing old fast) -------
bzs@BU-CS.BU.EDU (Barry Shein) (10/10/86)
You can certainly do backups to disk, you just shift your risks and costs. Note that disks are much more subject to various failures than tapes (not that tapes are perfect, not by any means.) One major reason for this is that your disk is constantly spinning and accessible (unless you turn it off and on which can make things worse, dismounting it helps of course, but not at all with mechanical failure.) Your tapes can be locked away someplace, safe from power surges, software errors and head crashes (saf*er* anyhow.) One possible compromise is to automate backups to a spare disk (assuming they really fit) and then backing *that* disk up to tape in one fell swoop which might be more efficient than running tapes and operators all over the place to different systems (this is particularly true if you could run the dumps over a network to the local disk.) It really all comes down to a systems analysis, how important are your backups to you? How disasterous would it be if you lost one? If you lost a series of them (all on one disk)? How much data are you backing up? Does it fit on one disk? How long do you keep backups? What are you promising your user's as far as file retrieval etc goes? There's no "answer", it's a complicated question. I keep hoping the "answer" will be inexpensive CD technology which should be the best of both worlds (on-line, operatorless, fairly large and robust, inexpensive*, removeable media.) I keep waiting though. -Barry Shein, Boston University * Not yet, this is the major problem.