Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!think.com!sdd.hp.com!uakari.primate.wisc.edu!aplcen!boingo.med.jhu.edu!haven.umd.edu!umbc3.umbc.edu!umbc4.umbc.edu!brian
From: brian@umbc4.umbc.edu (Brian Cuthie)
Newsgroups: comp.sys.next
Subject: Re: Disk transfer rates
Message-ID: <1991Jun13.191152.22357@umbc3.umbc.edu>
Date: 13 Jun 91 19:11:52 GMT
References: <64618@bbn.BBN.COM>
Sender: newspost@umbc3.umbc.edu (News  posting account)
Distribution: usa
Organization: Univ. of Maryland Baltimore County, Academic Computing Services
Lines: 66

In article <64618@bbn.BBN.COM> fwhite@BBN.COM () writes:
>
> I've been able to glean the following measurements of disk speed on
>an 040 NeXT from back issues of comp.sys.next.  I was wondering if
>anyone has the corresponding numbers for the Maxtor 8380S and 8780S
>and the Seagate ST1480?  I'm trying to narrow down what disk to buy.
>Thanks!
>
>Disk                            Transfer rate (read)
>--------------------		--------------------
>Micropolis 1588-15 		1.2MB/s
>QUANTUM LP105S			926172 bytes/s
>FUJITSU M2263S-512		841499 bytes/s		(SCSI-2 mode)
>FUJITSU M2263S-512		771374 bytes/s
>
>The benchmark:
>
>sylvester> sudo disk /dev/rsd0a
>disk name: QUANTUM LP105S 910109405
>disk type: fixed_rw_scsi
>Disk utility
>
>disk> read
>starting block? 0
># sectors per transfer? 16
>number of transfers? 1000
>sector increment? 16
>16384000 bytes in 17690 ms = 926172 bytes/s
>disk> quit
>
>Fred White            Bolt, Beranek, and Newman, Inc.
>fwhite@bbn.com	      150 CambridgePark Drive
>617 873-3536          Cambridge, MA 02140
>617 873-8202 (fax)


This is not exactly a good way to benchmark disks.  One real problem is that
the Quantum drives have a track cache that causes them to appear much 
faster than they are in actual use.

I did this once:  mounted a quantum Pro425 and a Seagate ST4385N.  I built
a duplicate system on the seagate from the quantum.  I then changed 
my default directory to the root of the Quantum's file system and 
entered the following command: "find . -name brian -print".  Needless
to say this caused all kinds of disk activity. 

When the command completed, unsuccesfully, I did an "iostat" and noticed that 
the Quantum has a real access time of around 20-21 ms.  This is a far cry
from the advertised 12 ms.  Mostly it's the difference between sequencial
and random reads.

I repeated the above steps with the Seagate and found that it averaged
in at about 13 ms.  Even this was above the 10.7 ms advertised by the
manufacturer.

My experience has been that drive specs are like mips.  Almost everyone
has a different way of measuring the performance of disk drives.  And,
as you would expect, companies pick the method that favors their
product.  Obviously, Quantum does sequencial reads.

My only real problem with the Imprimis [Seagate] drives, as I have posted here 
before, is that they make an inordinate amount of noise.  Otherwise, they are 
some of the best and fastest drives around. 


-brian