Path: utzoo!utgpu!watserv1!watmath!att!occrsh!uokmax!apple!usc!zaphod.mps.ohio-state.edu!ub!acsu.buffalo.edu
From: pjg@acsu.buffalo.edu (Paul Graham)
Newsgroups: comp.unix.large
Subject: Re: tape/backup systems
Message-ID: <35676@eerie.acsu.Buffalo.EDU>
Date: 12 Sep 90 05:48:12 GMT
References: <12701@encore.Encore.COM>
Sender: news@acsu.Buffalo.EDU
Organization: University at Buffalo
Lines: 23
Nntp-Posting-Host: urth.acsu.buffalo.edu

terryk@encore.com (Terence Kelleher) writes:
|How do large systems typically deal with the problem of backups?  I am
|trying to find answers for a 50-100 Gbyte disk system.  

well we use 8mm.  i don't have too awful a feeling about getting the
data off a year later.  epoch, qstar and some others are into optical
jukeboxes.  the epoch system is an automagic staging system that uses
magnetic and optical disk with an 8mm tape.  you turn it on and stuff
migrates around of its own volition.  neat -- pricey.  general atomics
has a variant of their mainframe system for unix.  it uses a dedicated
workstation with a fancy tape (maybe ampex) that holds 10-100 gig. or
something like that.

somebody (maybe legato) sells a smart network backup system.  how
smart?  well everyone shouts at the tape at the same time and despite
that the software keeps it straight (they say).  one presumes the idea
is to keep that 8mm streaming so you really can write 1G/hr/drive.

we had a little chat about this at the anaheim usenix.  (about 100
folks in a room for an hour).  rob kolstad moderated and posted the
outcome a bit ago.  look (or ask) in comp.org.usenix.
-- 
pjg@acsu.buffalo.edu / rutgers!ub!pjg / pjg@ubvms