NetBSD-Users archive
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]
Very slow file write/create operations
Hi,
I have 11.0RC2 installed on a HP ProLiant DL360p G8, with 128GB RAM and
RAID-1 via ciss(4).
# dmesg | grep ciss0
[ 1.020058] ciss0 at pci5 dev 0 function 0: HP Smart Array 12
[ 1.020058] ciss0: interrupting at msix4 vec 0
[ 1.020058] ciss0: 1 LD, HW rev 1, FW 8.00/8.00, 64bit fifo rro,
method perf 0x20000005
[ 1.020058] scsibus0 at ciss0: 1 target, 1 lun per target
# bioctl ciss0 show
Volume Status Size Device/Label Level Stripe
=============================================================
0 Online 466G CISS RAID 1 N/A
0:0 Online 466G 0:0.0 noencl <ATA TOSHIBA DT01ACA0>
0:1 Online 466G 0:1.0 noencl <ATA TOSHIBA DT01ACA0>
Already during installation I noticed that unpacking the archives takes
extremely long (hours). So I made a simple performance test with "cp"
and "rm" on the /etc directory:
# time cp -Rp /etc /root/
20.20 real 0.00 user 0.06 sys
# time rm -rf /root/etc
13.17 real 0.00 user 0.04 sys
Which is strange in two ways:
1. This is really slow! My older server (DL360 G5) did both in
Milliseconds!
2. Shouldn't the user- and sys-times more or less add into "real"? Where
did the system waste all that time?
On the other hand, creating a single 100MB file is quick as expected:
# dd if=/dev/zero of=/root/dummy bs=1m count=100
100+0 records in
100+0 records out
104857600 bytes transferred in 0.116 secs (903944827 bytes/sec)
Also reading is fast and writing to tmpfs:
# time cp -Rp /etc /tmp/
0.03 real 0.00 user 0.03 sys
# time rm -rf /tmp/etc
0.02 real 0.00 user 0.02 sys
I also compiled benchmarks/blogbench from pkgsrc, which should simulate
a busy file server. It gave the following result:
# blogbench -d /root/testdir
...
Final score for writes: 16
Final score for reads : 227925
Writes are more than factor 100000 slower!?
For comparison my old server had:
Final score for writes: 600
Final score for reads : 68988
Slower writes as well, but only factor 100.
What is wrong here? Does anybody have an idea what to check? Is it the
RAID or the ciss(4) driver? I remember I had similar results with 10.1.
The old server is running 8.3.
--
Frank Wille
Home |
Main Index |
Thread Index |
Old Index