From: Jan Kalcic (jandot_at_googlemail.com)
Date: Wed Aug 08 2007 - 10:14:44 CEST
Message-ID: <46B97B74.2070700@googlemail.com> Date: Wed, 08 Aug 2007 10:14:44 +0200 From: Jan Kalcic <jandot@googlemail.com> Subject: Re: [suse-sles-e] Slow direct I/O performance SLES 10 SP1
Andreas Mock wrote:
>> -----Ursprüngliche Nachricht-----
>> Von: "Gordon Ross" <G.Ross@ccw.gov.uk>
>> Gesendet: 06.08.07 19:57:41
>> An: <suse-sles-e@suse.com>
>> Betreff: Re: [suse-sles-e] Slow direct I/O performance SLES 10 SP1
>>
>> hq-hls3:~ # dd oflag=direct if=/dev/zero of=data.out bs=4k count=10000
>> 10000+0 records in
>> 10000+0 records out
>> 40960000 bytes (41 MB) copied, 0.88554 seconds, 46.3 MB/s
>>
>> HP DL385G2 on internal HD controller. (SATA discs)
>>
>>
>>
>> hq-hls1:~ # dd oflag=direct if=/dev/zero of=data.out bs=4k count=10000
>> 10000+0 records in
>> 10000+0 records out
>> 40960000 bytes (41 MB) copied, 41.4103 seconds, 989 kB/s
>>
>> HP DL385 on internal HD controller (SCSI HDs)
>>
>
> Hi Gordon, hi all,
>
> thanks to the guys who made the test for me. It's interesting.
> Gordon, could you make the test with a much bigger block count?
> Isn't it really strange that you get 1MB/s on one server and
> about 46.3 MB/s on the other. There must be some write buffering
> on the driver side. The 1MB/s on the second server is about
> what I get here.
>
> Best regards
> Andreas Mock
> _______________________________________________________________________
> Jetzt neu! Schützen Sie Ihren PC mit McAfee und WEB.DE. 3 Monate
> kostenlos testen. http://www.pc-sicherheit.web.de/startseite/?mc=022220
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: suse-sles-e-unsubscribe@suse.com
> For additional commands, e-mail: suse-sles-e-help@suse.com
>
>
>
As you can see, on SLES10 SP1 (1 SATA disks - software RAID 1) other
block are being written during the test whereas on openSuSE10.2 (1 PATA
disk, no RAID) there is no other writing activity. In fact, on the
server I have boring writing activity, even when the server is not
services anything. For what I could understand it's always buffering and
buffering and I don't get it at all.
Here is the output on SLES10 sp1 server:
#:~ # dd oflag=direct if=/dev/zero of=data.out bs=4k count=10000
10000+0 records in
10000+0 records out
40960000 bytes (41 MB) copied, 56.7699 seconds, 722 kB/s
#:~ # vmstat 1 8
procs -----------memory---------- ---swap-- -----io---- -system--
-----cpu------
r b swpd free buff cache si so bi bo in cs us sy
id wa st
1 1 164 333316 74656 2184888 0 0 8 8 9 8 9 15
66 11 0
1 1 164 333068 74852 2184692 0 0 0 516 1158 4200 21
31 4 43 0
1 1 164 333900 74036 2184480 0 0 0 552 1141 4256 20 34
10 35 0
1 1 164 333776 74216 2184300 0 0 4 568 1149 4161 22 32
25 21 0
1 1 164 333528 74432 2185112 0 0 0 556 1144 4111 19 34
12 35 0
1 1 164 333280 74648 2184896 0 0 0 560 1191 4261 20 34
22 24 0
1 1 164 333156 74848 2184696 0 0 0 532 1153 4212 23 31
20 26 0
2 1 164 332784 75048 2184496 0 0 0 676 1183 3903 19
66 8 7 0
Here is the output on openSuSE
#:~ # dd oflag=direct if=/dev/zero of=data.out bs=4k count=10000
10000+0 records in
10000+0 records out
40960000 bytes (41 MB) copied, 2.14537 s, 19.1 MB/s
#:~ # vmstat 1 8
procs -----------memory---------- ---swap-- -----io---- -system--
----cpu----
r b swpd free buff cache si so bi bo in cs us sy
id wa
0 0 0 339336 59828 465120 0 0 235 65 293 762 8 1
87 4
0 0 0 339336 59828 465120 0 0 0 0 253 224 1 0
99 0
0 0 0 339368 59828 465152 0 0 0 0 269 226 9 0
91 0
0 0 0 339368 59828 465152 0 0 0 0 252 202 0 0
100 0
0 0 0 339384 59828 465152 0 0 0 0 271 221 0 0
100 0
0 0 0 339384 59848 465132 0 0 0 20 255 212 0 0
99 1
0 0 0 339416 59848 465152 0 0 0 0 269 212 0 0
100 0
0 0 0 339416 59848 465152 0 0 0 0 251 208 0 0
100 0
---------------------------------------------------------------------
To unsubscribe, e-mail: suse-sles-e-unsubscribe@suse.com
For additional commands, e-mail: suse-sles-e-help@suse.com
This archive was generated by hypermail 2.1.7 : Wed Aug 08 2007 - 12:18:58 CEST