EXT4 vs XFS: large volumes with high-end RAID controller

Written by Gionatan Danti on . Posted in Linux & Unix

User Rating:  / 7
PoorBest 

PostgreSQL benchmarks:

Another popular, open source database server is PostgreSQL. Which filesystem is the fastest here?

The first test is about creating and populating a PostgreSQL database with 100 thousand rows, using sysbench oltp prepare test:

EXT4 vs XFS

XFS is slightly slower then EXT4 in this case.

As before, we want to test PostgreSQL performances both with simple, read-only queries:

EXT4 vs XFS

and complex (transactional, read-write) queries:

EXT4 vs XFS

Again the high-end controller helps XFS to remain in short distance from EXT4, but the latter is ahead by a ~10% margin.

Will pgbench confirm these data? Let's see the prepare time first:

EXT4 vs XFS

and execution scores then, expressed in transactions per second (TPS):

EXT4 vs XFS

While the prepare time was very similar, EXT4 remains the strong leader in TPS benchmark with a score almost 100% greater than XFS one.

So, it seems that for PostgreSQL databases, EXT4 remains (by far) the better choice.

Please note that pgbench transactional benchmark was the only one to be affected by write barriers on vs write barriers off: in the former case, I recorded performances over two times lower than in the latter (with both EXT4 and XFS).

Comments   

 
#1 Evgeny 2012-12-10 07:21
it's looks that you fsck time ext4 isn't true.
I think that you ran fsck.ext4 /dev/sd__something__
without "-f" key. It's means that fsck FS state
tune2fs -l
...
Filesystem state: clean
...
and if it clean do nothing
 

You have no rights to post comments