| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <4e5e476b0912081243ra071947if9a3edae77fb3861@mail.gmail.com>
Date: Tue, 8 Dec 2009 21:43:05 +0100
From: Corrado Zoccolo <czoccolo@...il.com>
To: Jeff Moyer <jmoyer@...hat.com>
Cc: Linux-Kernel <linux-kernel@...r.kernel.org>,
Jens Axboe <jens.axboe@...cle.com>,
Vivek Goyal <vgoyal@...hat.com>
Subject: Re: [PATCH] cfq-iosched: reduce write depth only if sync was delayed
On Tue, Dec 8, 2009 at 1:07 AM, Jeff Moyer <jmoyer@...hat.com> wrote:
> Corrado Zoccolo <czoccolo@...il.com> writes:
>
>> Hi Jeff, I remember you saw large performance drop on your SAN for
>> sequential writes with low_latency=1. Can you test if Shaohua's and
>> this patch fix allow to recover some bandwidth? I think that enabling
>> the queue depth ramp up only if a sync request was delayed should
>> disable it for fast hardware like yours, so you should not be seeing
>> the slowdown any more.
>
> Average of 10 runs. Low latency set to 0:
>
> Unit information
> ================
> File size = megabytes
> Blk Size = bytes
> Rate = megabytes per second
> CPU% = percentage of CPU used during the test
> Latency = milliseconds
> Lat% = percent of requests that took longer than X seconds
> CPU Eff = Rate divided by CPU% - throughput per cpu load
>
> Sequential Reads
> File Blk Num Avg Maximum Lat% Lat% CPU
> Identifier Size Size Thr Rate (CPU%) Latency Latency >2s >10s Eff
> ---------------------------- ------ ----- --- ------ ------ --------- ----------- -------- -------- -----
> 2.6.32 8192 65536 8 90.08 94.51% 51.316 10268.25 0.00000 0.00000 95
> 2.6.32 8192 65536 16 99.36 199.8% 89.248 13883.81 0.00000 0.00000 50
>
> Random Reads
> File Blk Num Avg Maximum Lat% Lat% CPU
> Identifier Size Size Thr Rate (CPU%) Latency Latency >2s >10s Eff
> ---------------------------- ------ ----- --- ------ ------ --------- ----------- -------- -------- -----
> 2.6.32 8192 65536 8 81.16 85.90% 45.672 5963.19 0.00000 0.00000 94
> 2.6.32 8192 65536 16 116.32 230.3% 58.371 6098.36 0.00000 0.00000 51
>
> Sequential Writes
> File Blk Num Avg Maximum Lat% Lat% CPU
> Identifier Size Size Thr Rate (CPU%) Latency Latency >2s >10s Eff
> ---------------------------- ------ ----- --- ------ ------ --------- ----------- -------- -------- -----
> 2.6.32 8192 65536 8 112.17 1085.% 42.623 17114.54 0.00152 0.00000 10
> 2.6.32 8192 65536 16 111.26 2117.% 84.964 26480.60 0.03202 0.00000 5
>
> Random Writes
> File Blk Num Avg Maximum Lat% Lat% CPU
> Identifier Size Size Thr Rate (CPU%) Latency Latency >2s >10s Eff
> ---------------------------- ------ ----- --- ------ ------ --------- ----------- -------- -------- -----
> 2.6.32 8192 65536 8 162.48 887.1% 6.106 313.54 0.00000 0.00000 18
> 2.6.32 8192 65536 16 156.38 1767.% 14.077 1254.62 0.00000 0.00000 9
>
>
> Average of 10 runs. Low latency set to 1:
>
> Unit information
> ================
> File size = megabytes
> Blk Size = bytes
> Rate = megabytes per second
> CPU% = percentage of CPU used during the test
> Latency = milliseconds
> Lat% = percent of requests that took longer than X seconds
> CPU Eff = Rate divided by CPU% - throughput per cpu load
>
> Sequential Reads
> File Blk Num Avg Maximum Lat% Lat% CPU
> Identifier Size Size Thr Rate (CPU%) Latency Latency >2s >10s Eff
> ---------------------------- ------ ----- --- ------ ------ --------- ----------- -------- -------- -----
> 2.6.32 8192 65536 8 92.23 100.1% 52.119 6920.18 0.00000 0.00000 92
> 2.6.32 8192 65536 16 97.88 217.0% 99.691 7453.18 0.00000 0.00000 45
>
> Random Reads
> File Blk Num Avg Maximum Lat% Lat% CPU
> Identifier Size Size Thr Rate (CPU%) Latency Latency >2s >10s Eff
> ---------------------------- ------ ----- --- ------ ------ --------- ----------- -------- -------- -----
> 2.6.32 8192 65536 8 98.70 107.5% 42.994 3409.08 0.00000 0.00000 92
> 2.6.32 8192 65536 16 140.41 323.9% 59.616 4525.46 0.00000 0.00000 43
>
> Sequential Writes
> File Blk Num Avg Maximum Lat% Lat% CPU
> Identifier Size Size Thr Rate (CPU%) Latency Latency >2s >10s Eff
> ---------------------------- ------ ----- --- ------ ------ --------- ----------- -------- -------- -----
> 2.6.32 8192 65536 8 112.33 1076.% 42.617 17072.28 0.00076 0.00000 10
> 2.6.32 8192 65536 16 111.84 2097.% 85.156 28221.77 0.02976 0.00000 5
>
> Random Writes
> File Blk Num Avg Maximum Lat% Lat% CPU
> Identifier Size Size Thr Rate (CPU%) Latency Latency >2s >10s Eff
> ---------------------------- ------ ----- --- ------ ------ --------- ----------- -------- -------- -----
> 2.6.32 8192 65536 8 159.33 870.5% 6.469 765.50 0.00000 0.00000 18
> 2.6.32 8192 65536 16 141.60 1632.% 15.364 2337.57 0.00000 0.00000 9
>
>
> Cheers,
> Jeff
>
The numbers look good. Now, there is no penalty in having low_latency
set for sequential writes, and just a small penalty for random ones.
The fact that random reads are faster with low_latency set is interesting.
Is the test is running with your patched tiobench (so that the number
of random operations is comparable with sequential ones)?
Thanks,
Corrado
--
__________________________________________________________________________
dott. Corrado Zoccolo mailto:czoccolo@...il.com
PhD - Department of Computer Science - University of Pisa, Italy
--------------------------------------------------------------------------
The self-confidence of a warrior is not the self-confidence of the average
man. The average man seeks certainty in the eyes of the onlooker and calls
that self-confidence. The warrior seeks impeccability in his own eyes and
calls that humbleness.
Tales of Power - C. Castaneda
Powered by blists - more mailing lists