Date: Tue, 9 Apr 1996 20:44:20 +0200 From: se@zpr.uni-koeln.de (Stefan Esser) To: Kees Jan Koster <dutchman@spase.nl> Cc: freebsd-hackers@freebsd.org (FreeBSD hackers Mailing list) Subject: Re: HDD cpu usage (IDE vs. SCSI). Message-ID: <199604091844.AA05369@Sisyphos> In-Reply-To: Kees Jan Koster <dutchman@spase.nl> "HDD cpu usage (IDE vs. SCSI)." (Apr 1, 11:03)
next in thread | previous in thread | raw e-mail | index | archive | help
On Apr 1, 11:03, Kees Jan Koster wrote:
} Subject: HDD cpu usage (IDE vs. SCSI).
} Hoi Hackers,
}
} I was told that cpu usage of SCSI disks was lower than that of IDE disks.
Yes, in most situations it definitely is,
especially given the lack of EIDE DMA
support in FreeBSD (not that I'd need it :)
} However:
[ indentation and unimportant lines removed ... ]
LikeEver (amd486DX4-100, ncr pci scsi controller)
-------Sequential Output-------- ---Sequential Input-- --Random--
-Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks---
Directory MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec %CPU
Seagate 20 1299 81.1 1263 15.5 653 18.2 1778 94.9 1982 41.8 47.7 5.7
Maxtor 20 750 41.9 740 8.5 405 9.7 764 34.7 770 8.8 32.0 3.9
phobos (pentium 90, on-board IDE controller)
Directory MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec %CPU
Quantum 40 1215 41.7 1818 18.8 811 11.5 1376 41.0 1990 16.0 39.9 3.9
} How come the seagate uses twice as much cpu as the quantum?
How comes the 486 needs twice as much time
as a P90 to complete just about any task ? ;-)
Honestly: I'm not sure what you are doing
in these tests, since the numbers are low
for the NCR SCSI as well as the IDE case.
I'd guess that both the Seagate and the
Maxtor are old drives (or they'd read more
than 2MB/s in the "Block Input" test).
How about a different test:
# time dd if=/dev/rsd0c of=/dev/null bs=64k count=1000
1000+0 records in
1000+0 records out
65536000 bytes transferred in 9 secs (7281777 bytes/sec)
9.32 real 0.02 user 0.24 sys
(ASUS SP3G with NCR and Quantum Atlas 2GB.)
What are your numbers for both the NCR and
the EIDE system ?
These numbers indicate less than 3% CPU
is required to read 7MB/s (though there
are cycles that aren't accounted to the
"dd" process and which aren't easily
measured).
I'd guess that less than 1% CPU is spent
on doing the actual transfer on you amd486
system. The remaining CPU time is consumed
by file system overhead.
Regards, STefan
--
Stefan Esser, Zentrum fuer Paralleles Rechnen Tel: +49 221 4706021
Universitaet zu Koeln, Weyertal 80, 50931 Koeln FAX: +49 221 4705160
==============================================================================
http://www.zpr.uni-koeln.de/~se <se@ZPR.Uni-Koeln.DE>
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199604091844.AA05369>