Subject: Re: 160 GB HD on PWS
To: None <port-alpha@netbsd.org>
From: Ray Phillips <r.phillips@jkmrc.uq.edu.au>
List: port-alpha
Date: 06/06/2003 15:02:24
>Can you boot from that disk?
Yes,
>>>show dev
dkb0.0.1.4.0 DKB0 ST316002 1A 3.06
dkc0.0.0.1010.0 DKC0 RZ2CC-KB DC2B
dkc100.1.0.1010.0 DKC100 SEAGATE ST318436LW 0005
dkc200.2.0.1010.0 DKC200 SEAGATE ST318417N 0105
dkc500.5.0.1010.0 DKC500 PLEXTOR CD-R PX-W1210S 1.06
dva0.0.0.0.1 DVA0
ewa0.0.0.3.0 EWA0 00-00-F8-75-5E-6F
pkc0.7.0.1010.0 PKC0 SCSI Bus ID 7 5.57
pqa0.0.0.4.0 PQA0 PCI EIDE
pqb0.0.1.4.0 PQB0 PCI EIDE
>>>
"boot dkb0" worked fine.
>Bechmarks? (bonnie?)
>What CPU load do you get on heavy IDE activity compared to SCSI?
I only had access to the disk for one night and didn't run a real
benchmark using it. The best I can offer is the time it took to copy
files to it from two different partitions on SCSI disks. The first
source partition I used was on a ccd volume spread across dkc100 and
dkc200 (I thought I'd make a 50 GB file for fun). The SCSI
controller was a QLogic ISP1040B.
isp0 at pci1 dev 10 function 0: QLogic 1020 Ultra Wide SCSI HBA
isp0: interrupting at dec 550 irq 20
scsibus0 at isp0: 16 targets, 8 luns per target
pws500# ls -l
total 6817458
-rw-r--r-- 1 root wheel 1073741824 Mar 28 17:06 1-GB-file
-rw-r--r-- 1 ray wheel 5368709120 Mar 26 02:29 5-GB-file
-rw-r--r-- 1 root wheel 536870912 Jun 2 20:53 512-MB-file
-rw-r--r-- 1 ray wheel 335 Mar 29 10:43 p.c
pws500# /usr/bin/time cp 5-* /mnt/big*
384.77 real 0.19 user 137.30 sys
pws500# /usr/bin/time cat 5-*>>/mnt/big*/*
383.37 real 0.75 user 155.97 sys
pws500# sh
# ls -l /mnt/big*
total 10490888
-rw-r--r-- 1 root wheel 10737418240 Jun 3 01:05 5-GB-file
# for i in 3 4 5 6 7 8 9 10
> do
> /usr/bin/time cat 5-*>>/mnt/big-file/5-GB-file
> done
385.17 real 0.62 user 156.62 sys
385.79 real 0.76 user 154.76 sys
387.61 real 0.81 user 159.38 sys
385.15 real 0.65 user 159.89 sys
388.79 real 0.76 user 155.33 sys
393.55 real 0.69 user 160.51 sys
400.23 real 0.76 user 160.21 sys
397.02 real 0.77 user 155.28 sys
# ls -l /mnt/big*
total 52454424
-rw-r--r-- 1 root wheel 53687091200 Jun 3 02:12 5-GB-file
# wc -l /mnt/big*/*
838860800 /mnt/big-file/5-GB-file
# head -1 /mnt/big*/*
012345678901234567890123456789012345678901234567890123456789012
# head -1 /mnt/big*/* | wc -c
64
#
So that's a copy rate of about 13 MB/s.
The second source partition was on dkc0:
pws500# ls -l
total 1049096
-rw-r--r-- 1 root wheel 1073741824 Jun 3 03:49 1-GB-file
pws500# sh
# for i in 1 2
> do
> /usr/bin/time cp * /mnt/big*
> done
113.46 real 0.03 user 25.43 sys
114.10 real 0.03 user 24.94 sys
#
So, about 9 MB/s.
Copying from the ccd volume to dkc0 only achieved ~ 4 MB/s:
# ls -l
total 6817458
-rw-r--r-- 1 root wheel 1073741824 Mar 28 17:06 1-GB-file
-rw-r--r-- 1 ray wheel 5368709120 Mar 26 02:29 5-GB-file
-rw-r--r-- 1 root wheel 536870912 Jun 2 20:53 512-MB-file
-rw-r--r-- 1 ray wheel 335 Mar 29 10:43 p.c
# for i in 1 2 3 4 5
> do
> /usr/bin/time cp 1-* /usr/cp-tst
> done
250.39 real 0.03 user 24.92 sys
252.69 real 0.03 user 24.89 sys
252.92 real 0.03 user 25.23 sys
252.62 real 0.02 user 24.91 sys
252.64 real 0.03 user 24.87 sys
#
Ray