Hi folks,
Some days ago I asked about what cure would apply to a slow harddisk.
Well, actually our dealer had supplied the wrong disk (|) -- this indeed
was the reason it was so slow. Nevertheless I received three responses
which may prove helpful to you too. Thanks to all who took the time to
repond. First comes my original question, then the answers:
----------------------------Original message----------------------------
Hi folks,
Having exchanged my 'old' IPC for an ELC (8MB) I also changed my
harddisk. I how have a Seagate ST4767NV ('runner') which is suspected
to have an average access time of about 11msec.
Well what really is the case is, that the new disk is VERY SLOOOOOWWW.
Copying a 1.5MB-file takes 15 seconds -- with no logged in users,
nearly no daemons running etc. This simply can't be normal|
With my old disk this took about 1-2secs.
I suspect that something with my format.dat is wrong, i.e. I'm treating
the disk like 'another one'. The ST4767NV is formatted as a "SUN0669"
with the format.dat as below:
partition = "SUN0669" \
: disk = "SUN0669" : ctlr = MD21 \
: a = 0, 32400 : b = 40, 64800 : c = 0, 1307340 : g = 120, 1210140
A "probe-scsi" tells me that the disk actually is a
'CDC 94191-15'.
Any glue what's wrong with this setup?
I would really appreciate a working format.dat, i.e. one that makes the
disk a real '11msec' one.
Thanks very much in advance for any help,
-ewald
--
Ewald Jenisch | E-mail: jenisch@univie.ac.at (Internet)
Vienna University Computer Center | z00ejr01@awiuni11.bitnet (Bitnet)
Universitaetsstrasse 7 | Tel: +43 (222) 43-61-11 x251 Fax x170
A-1010 Vienna, Austria | NIC-Handle: EJ51
----------------------------Original message----------------------------
Another possibility is that the cable is off - if it has to retry the
xfers all the time, the disk seems pretty slow. Check your termination
and cable.
l & h,
kev
Kevin Sheehan
Optimation Software Engineering
kevin.sheehan@fourx.aus.sun.com
----------------------------Original message----------------------------
With SCSI, format.dat shouldn't affect performance since it is pretty much
addressed as logical sectors rather than cylinders/heads/sectors. Assuming
that format and label go OK, you are not exceeding the bounds of the drive.
Access time is not always the key measure of performance either. Transfer rate
is just as important (more so for large files). This is solely determined by
the disk geometry itself, ie the number of sectors/track, since this defines
just how fast the data spins under the heads.
The disk with the higher density, will get better transfer rates.
The large difference you see though sounds like something more fundamental,
possibly. Make sure the SCSI bus is properly terminated, and the cables
arent too long.
Russ Poffenberger DOMAIN: poffen@sj.ate.slb.com
Schlumberger Technologies UUCP: {uunet,decwrl,amdahl}!sjsca4!poffen
1601 Technology Drive CIS: 72401,276
San Jose, Ca. 95110 Voice: (408)437-5254 FAX: (408)437-5246
----------------------------Original message----------------------------
I've found that there are a couple of parameters which greatly affect the
throughput you get on a disk drive -- primarily the rotational delay,
followed by "maxcontig".
You can get the current value of both parameters with dumpfs.
I've found that about 4ms seems to be best rotational delay value for most
SCSI disks; in most cases, you will get 0ms by default, which is must worse.
You can use "tunefs" to change it, with the partition dismounted.
What I usually do with a new drive is to try a number of different values
until I get the one I like best. I just start from zero, work up until
things are definitely getting worse. To measure, I run iostat in one
window, while using dd to exercise the drive in another. Something like:
dd if=/dev/zero of=/newdisk/something bs=16k
Maxcontig defaults to 7, I believe, which is usually about best. You can
always try adjusting it, though.
Jonathan Corbet
National Center for Atmospheric Research, Atmospheric Technology Division
corbet@stout.atd.ucar.edu
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:06:17 CDT