Re: [tlug] SSD caching of 10x 10GB files on slow NAS

[Miles Colman <mcolman@example.com>]

Holy Old Thread Batman!

Now that SSDs with fast PCI connects are going mass-market, I'm thinking about looking at this again. If you were going to choose between learning memcached  or learning gnu make to copy things to shm similar to this post: http://widgetsandshit.com/teddziuba/2011/02/stupid-unix-tricks-workflow-control-with-gnu-make.html

which would you do?

On Thu, Jan 3, 2013 at 9:55 AM, Kalin KOZHUHAROV <me.kalin@example.com> wrote:

On Sat, Dec 29, 2012 at 2:17 AM,  <jep200404@example.com> wrote:
> On Fri, 28 Dec 2012 23:43:32 +0900, Kalin KOZHUHAROV <me.kalin@example.com> wrote:
>
>> On Fri, Dec 28, 2012 at 10:45 PM,  <jep200404@example.com> wrote:
>> > On Fri, 28 Dec 2012 12:57:55 +0900, Kalin KOZHUHAROV <me.kalin@example.com> wrote:
>> >
>> >> 960GB SSD with 1.5GB/s R/W
>> >
>> > 1.5 Gigabytes/s is smoking! Did you mean 1.5 Gigabits/s?
>
>> Yes, it is smoking:
>> http://www.ocztechnology.com/ocz-revodrive-3-x2-pci-express-ssd.html
>> 1.5GB/s read, 1.3GB/s write, 230k iops
>>
>> So, 960GB drive is read into /dev/null in about 11 minutes ...
>
> Wow. That is _seriously_ smoking. Awesome!
>
> I can see how it might be hard for the rest of the computer to
> keep up. It'll be interesting to see the output of the following
> (with sda modified for the correct drive) when you get around to it.
>
> time dd if=/dev/sda of=/dev/null bs=1M
>

$ echo 3 > /proc/sys/vm/drop_caches; sync; /usr/bin/time dd
if=/dev/md127 of=/dev/zero bs=1M count=100000
100000+0 records in
100000+0 records out
104857600000 bytes (105 GB) copied, 127.346 s, 823 MB/s
0.02user 33.36system 2:07.34elapsed 26%CPU (0avgtext+0avgdata 7536maxresident)k
0inputs+0outputs (1major+529minor)pagefaults 0swaps

$ echo 3 > /proc/sys/vm/drop_caches; sync; /usr/bin/time dd
if=/dev/zero of=/mnt/pictures/100G bs=1M count=100000; /usr/bin/time
sync
100000+0 records in
100000+0 records out
104857600000 bytes (105 GB) copied, 97.4376 s, 1.1 GB/s
0.05user 58.80system 1:37.45elapsed 60%CPU (0avgtext+0avgdata 7536maxresident)k
0inputs+0outputs (3major+527minor)pagefaults 0swaps
0.00user 0.00system 0:05.41elapsed 0%CPU (0avgtext+0avgdata 2464maxresident)k
0inputs+0outputs (1major+203minor)pagefaults 0swaps
$ ls -lsh /mnt/pictures/100G
98G -rw-r--r-- 1 root root 98G Jan  4 02:47 /mnt/pictures/100G

Well, just half of "advertised"... I'll see if it can be improved.
Beats me why writing is faster (even with added sync).

Anyway, that was just a quick test, I'll report more when I start
using it more and I hope to get it run faster. For now I am happy it
runs at all (there was bug 51881, so linux-3.7.1 was not working,
fixed now).

Cheers,
Kalin.

--
To unsubscribe from this mailing list,
please see the instructions at http://lists.tlug.jp/list.html

The TLUG mailing list is hosted by ASAHI Net, provider of mobile and
fixed broadband Internet services to individuals and corporations.
Visit ASAHI Net's English-language Web page: http://asahi-net.jp/en/