Re: Comments requested on IO performance : new db server

From: Tomas Vondra <tv(at)fuzzy(dot)cz>
To: pgsql-performance(at)postgresql(dot)org
Subject: Re: Comments requested on IO performance : new db server
Date: 2012-03-10 15:12:39
Message-ID: 4F5B6F67.1040300@fuzzy.cz
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-performance

On 10.3.2012 11:51, Rory Campbell-Lange wrote:
> Is a block size of 4096 a good idea both for the filesystem and
> postgresql? The analysis here:
> http://www.fuzzy.cz/en/articles/benchmark-results-hdd-read-write-pgbench/
> appears to suggest that at least for database block sizes of 4096
> read/write performance is much higher than for smaller block sizes.

Hi,

interpreting those results is a bit tricky for several reasons. First,
those are 'average results' for all filesystems (and the behavior of
filesystems may vary significantly). I'd recommend checking results for
the filesystem you're going to use (http://www.fuzzy.cz/bench)

Second, the article discusses just TPC-B (OLTP-like) workload results.
It's quite probable your workload is going to mix that with other
workload types (e.g. DSS/DWH). And that's exactly where larger block
sizes are better.

To me, 8kB seems like a good compromise. Don't use other block sizes
unless you actually test the benefits for your workload.

Tomas

>
> Rory
>
> On 09/03/12, Rory Campbell-Lange (rory(at)campbell-lange(dot)net) wrote:
>> ...An ancillary question is whether a 4096 block size is a good idea.
>> I suppose we will be using XFS which I understand has a default block
>> size of 4096 bytes.
>>
>> RAID 10
>> --------------------------------------
>> Read sequential
>>
>> BS MB/s IOPs
>> 512 0129.26 264730.80
>> 1024 0229.75 235273.40
>> 4096 0363.14 092965.50
>> 16384 0475.02 030401.50
>> 65536 0472.79 007564.65
>> 131072 0428.15 003425.20
>> --------------------------------------
>> Write sequential
>>
>> BS MB/s IOPs
>> 512 0036.08 073908.00
>> 1024 0065.61 067192.60
>> 4096 0170.15 043560.40
>> 16384 0219.80 014067.57
>> 65536 0240.05 003840.91
>> 131072 0243.96 001951.74
>> --------------------------------------
>> Random read
>>
>> BS MB/s IOPs
>> 512 0001.50 003077.20
>> 1024 0002.91 002981.40
>> 4096 0011.59 002968.30
>> 16384 0044.50 002848.28
>> 65536 0156.96 002511.41
>> 131072 0170.65 001365.25
>> --------------------------------------
>> Random write
>>
>> BS MB/s IOPs
>> 512 0000.53 001103.60
>> 1024 0001.15 001179.20
>> 4096 0004.43 001135.30
>> 16384 0017.61 001127.56
>> 65536 0061.39 000982.39
>> 131072 0079.27 000634.16
>> --------------------------------------

In response to

Browse pgsql-performance by date

  From Date Subject
Next Message McGehee, Robert 2012-03-11 17:38:04 Feature Request: No pg_dump lock on unlogged tables
Previous Message Rory Campbell-Lange 2012-03-10 10:51:12 Re: Comments requested on IO performance : new db server