Article View: misc.wanted
Article #3048Re: Multivendor network backup for unix workstations
From: Marcus J. "will do TCP/IP for fo...
Date: Tue, 19 May 1992 02:02
Date: Tue, 19 May 1992 02:02
75 lines
5016 bytes
5016 bytes
rmilner@zia.aoc.nrao.edu (Ruth Milner) writes: >So, before I go call Legato and order it :-), can you tell us a bit more: > > - how much (just roughly) did it cost? > - did you have to pay by the number of systems being backed up, the > number of tape drives you will use with it, or just a single licence, > say for the system you will be initiating all of this from? The DECNsr product (built on Legato's networker) is licensed based on one server and then a smaller fee for each client. It comes out to about $2k for the server (you need one and maybe 50-60Mb of disk for the indices) and about $300 per client. There's no charge for the tape device support, so you can support a mess 'o DATs if you have 'em. For exact pricing look at your DECDirect catalog. Clients for MIPS and Sun and HP and of course ULTRIX are available presently with more in the works. Let me digress for a moment on the topic of backups (not just to add some technical stuff so people don't complain that I'm acting like a salesrep) (incidentally, if you accuse me of acting like a sales rep, I choose swords at dawn, have your seconds meet mine at USENIX and we'll settle the matter like gentlebeings) *ahem* anyhow - I've noticed that tools like NSR or Networker should cause users to shift their backup paradigms. A lot of folks have come to me asking for a network backup solution that can handle a "weekly backup of 20Gb" or something like that. The day in which it was necessary to do full weeklies, I submit, it past. With the quick seeking ability and small form factor of a DAT, it's perfectly reasonable to do bimonthlies, in my opinion. As long as the indexes are kept, you can recover a whole file system just about as fast as having it on one reel used to be. Since the tape archive server knows what went where, I've noticed that recovering stoff a year old out of my tapes seldom requires mounting more than one tape, and has never yet needed more than two. On my machines I have been doing nothing but incrementals for over a year and a half. I haven't ever recycled a tape. I now have 16 of them. I can recover quickly and precisely any rev of any file in a matter of minutes. This strikes some people as utterly insane, since if I lose a whole disk it'll take me longer to recover, but in fact I haven't done that yet (lost a whole disk) and I'm pretty sure that if I did I'd be able to probably retrieve the whole filesystem in an hour or two. Not too terrible, really. Considering that with incremental backups only, I have to change tapes about every month or so, the administrative load is very light. This also means I don't even NEED a changer. Typically if I need to recover something, the odds are high that it's on the tape already in the drive. Has anyone else experimented with changing their outlook on backups? My thinking went as follows: a) I used to like to have a monthly level 0 dump 'cuz I knew that if I wanted a file it'd be on that tape and I wouldn't have to grovel in bins for it b) if I lose a whole disk I will be able to recover in 2 operations (1 - recover the level 0 and, 2 - add the latest level 5 and maybe 9) c) waitaminnit! if I have an index that lets me KNOW where my file is stored (and it'll just *tell* me "you dope, mount tape #3") item 'a' above is pointless d) if my full disk recovery plan in 'b' above requires 2 steps, maybe 3, that's not a whole lot better than just shlunking a few DATs e) one royal PAIN with dump is that level 0 dumps don't know when a file has been *deleted* since the dump. if my tool's index lets me say "recover this directory as it looked on Jan 2, 1992" I am actually better off and less likely to have to shlunk DATs. Comments? mjr.
Technical Details
Message-ID:
<1992May19.020223.16714@decuac.dec.com>Path:
arc1-ca.pugleaf.net!sparky!uunet!haven.umd.edu!decuac!hussar.dco.dec.com!mjrReferences:
<1992May15.175312.3592@watdragon.waterloo.edu> <1992May16.000918.7220@odin.corp.sgi.com> <1992May18.230822.1368@zia.aoc.nrao.edu>Show Full Headers
Complete Article Headers
X-Google-Language: ENGLISH,ASCII-7-bit X-Google-Thread: f48e5,db8056cc53e3f22 X-Google-Attributes: gidf48e5,public X-Google-Thread: 10d023,db8056cc53e3f22 X-Google-Attributes: gid10d023,public X-Google-Thread: f9532,db8056cc53e3f22 X-Google-Attributes: gidf9532,public X-Google-Thread: f8ff0,db8056cc53e3f22 X-Google-Attributes: gidf8ff0,public X-Google-Thread: 1164f8,db8056cc53e3f22 X-Google-Attributes: gid1164f8,public X-Google-ArrivalTime: 1992-05-18 14:13:06 PST Newsgroups: misc.wanted,aus.wanted,comp.sys.sgi,comp.sys.dec,comp.sys.sun.wanted Path: arc1-ca.pugleaf.net!sparky!uunet!haven.umd.edu!decuac!hussar.dco.dec.com!mjr From: mjr@hussar.dco.dec.com (Marcus J. "will do TCP/IP for food" Ranum) Subject: Re: Multivendor network backup for unix workstations Message-ID: <1992May19.020223.16714@decuac.dec.com> Sender: news@decuac.dec.com (USENET News System) Nntp-Posting-Host: hussar.dco.dec.com Organization: Digital Equipment Corporation, Washington ULTRIX Resource Center References: <1992May15.175312.3592@watdragon.waterloo.edu> <1992May16.000918.7220@odin.corp.sgi.com> <1992May18.230822.1368@zia.aoc.nrao.edu> Date: Tue, 19 May 1992 02:02:23 GMT Lines: 75