What do u make of this btrfs/ext4 benchmark? Is the benchmark correct?

[latebeat]

Hi all,

I recently bought a 2nd ssd drive for my laptop in order to use the built in (fake) controller from intel and setup a raid-0 array.
So I did and I'm trying to take some meaningfull benchmarks as to which filesystem will work best for my setup.
I've build a custom suite basted on pts/disk but I've taken out all the database related tests.
Finally, I wanted to see the impact of full disk encryption under different file systems.

Now when it comes to reads, I didn't see much more than a 10% deviation between the filesystems.


But... what buffles me is btrfs seems to be WAY faster than all the other systems, when it comes to writes, to the point that it doesn't seem right!
In fact, if you see the screenshots below, it's so much faster than ext4 that even with full disk encryption (dmcrypt) it's still MUCH faster than unecrypted ext4, so I thought, that can't be right?

What do u guys make of these results? Does it make sense?



-----------------------------------
Here are the screenshots:













Also another thing is that when it comes to random writes the performance plummets vertically...

[curaga]

Weren't those tests writing zeroes, which tend to compress well (your lzo option)?


(raid0! With a bad hw controller! It's like you're begging to lose your data...)

[latebeat]

Weren't those tests writing zeroes, which tend to compress well (your lzo option)?


(raid0! With a bad hw controller! It's like you're begging to lose your data...)

I don't know, I'm pretty new to the test suite that's why I'm asking?