NetNews Usenet Archive 1992 #27/NN_1992_27.iso/spool/comp/os/vms/18072

Internet Message Format | 1992-11-17 | 1.7 KB

Path: sparky!uunet!ogicse!uwm.edu!zaphod.mps.ohio-state.edu!moe.ksu.ksu.edu!kuhub.cc.ukans.edu!sloane
From: [e-mail] (Bob Sloane)
Newsgroups: comp.os.vms
Subject: Re: ?Why does this backup go so slow?
Message-ID: <[e-mail]>
Date: 17 Nov 92 15:42:27 GMT
Article-I.D.: kuhub.1992Nov17.094227.44871
References: <[e-mail]>
Organization: University of Kansas Academic Computing Services
Lines: 22

In <[e-mail]>, [e-mail] writes:
> It took us 4.5 hours to do an image backup of an RA90 (80% full of user
> data) onto an RA92?  Why does it not go faster?   [I had prepared a much
> longer version of the question, with so many particulars, that my boss
> said to just ask the simple question.]

This is just a guess.  An RA90 holds about 2376153 blocks. If the
device was 80% full, then you had to copy about 780,000,000 bytes.
That means your transfer rate was 780,000,000/(4.5*60*60)=48148 bytes
per second.  Our RA90s here seem to max out about 40 I/Os per second,
or 25ms access time.  If each I/O that backup does averages 1200
bytes(48,000/40), then the I/O time on the RA90 would make sense. Is
the cluster factor on the disk 3? Back may do its I/O based on the
cluster size of the disk.  If the disk is fragmented, that might also
slow down the I/O.

If both devices are connected via the same channel, then you have to
consider the bandwidth of the channel. Will it support more then 100k
bytes per second?
--
USmail: Bob Sloane, University of Kansas Computer Center, Lawrence, KS, 66045
E-mail: [e-mail], [e-mail], AT&T: [phone]