On Mon, May 8, 2017 at 4:24 PM, Vincent Veyron <vv.li...@wanadoo.fr> wrote:
> On Mon, 8 May 2017 12:48:29 -0600
> Scott Marlowe <scott.marl...@gmail.com> wrote:
>
> Hi Scott,
>
> Thank you for your input.
>
>>
>> The most likely cause of the difference would be that one server IS
>> honoring fsync requests from the db and the other one isn't.
>>
>> If you run pgbench on both (something simple like pgbench -c 1 -T 60,
>> aka one thread for 60 seconds) on a machine running on a 7200RPM hard
>> drive, you should get approximately 120 transactions per second
>
> Here are the results :
>
> #Kimsufi
> pgbench -c 1 -T 60 test
> starting vacuum...end.
> transaction type: TPC-B (sort of)
> scaling factor: 1
> query mode: simple
> number of clients: 1
> number of threads: 1
> duration: 60 s
> number of transactions actually processed: 6618
> latency average: 9.069 ms
> tps = 110.270771 (including connections establishing)
> tps = 110.283733 (excluding connections establishing)

Just under 120, looks like fsync is working.

>
> #Online
> starting vacuum...end.
> transaction type: TPC-B (sort of)
> scaling factor: 1
> query mode: simple
> number of clients: 1
> number of threads: 1
> duration: 60 s
> number of transactions actually processed: 1150
> latency average: 52.317 ms
> tps = 19.114403 (including connections establishing)
> tps = 19.115739 (excluding connections establishing)

OK that's horrendous. My mobile phone is likely faster. We need to
figure out why it's so slow. If it's in a RAID-1 set it might be
syncing.

>> > -Why are regular queries much faster on this same server?
>>
>> That's a whole nother subject. Most likely the faster machine can fit
>> the whole db in memory, or has much faster memory, or the whole
>> dataset is cached etc etc.
>>
>
> The dataset is small (35 MB) and both servers have 4GB memory. It appears to 
> be faster on the Online server.

Yeah it fits in memory. Select queries will only hit disk at bootup.

First machine
SNIP
>         Speed: 1066 MHz
SNIP
>         Configured Clock Speed: 1066 MHz

Second machine

>         Speed: 1600 MHz
SNIP
>         Configured Clock Speed: 1333 MHz

Yeah the second machine likely has a noticeably faster CPU than the
first as well. It's about two years younger so yeah it's probably just
cpu/mem that's making it fast.


-- 
Sent via pgsql-performance mailing list (pgsql-performance@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-performance

Reply via email to