A debate I had recently about the usage of reiser4 and the bad publicity it has been attracting. Please note that I am not an expert in filesystems, I just tested things and formed an opinion.
The whole thing might not be coherent, but the points are clear, and AFAIK true.
- I know reiser4 lags behind on some benchmarks.
- I did benchmarks on my computer.
- I didn't say reiser4 is the fastest filesystem. I don't care about those who do.
- I know it uses more RAM. That's called caching. XFS does that too.
- However the journal's overhead is lower than reiserfs-3.6 or ext3. ( not preallocated )
- As to using more CPU time, I'm not sure about this, I think it's because the online packer is currently disabled/not used, with time we get more fragmentation ( due to tighter packing of files ), which means more seeking and more overhead. The online packer is promised to be back in reiser4.1 and there is no offline packer ( except if you tar the whole filesystem somewhere else and then untar it ).
Some reasons to use reiser4 :
- reason 1: I live in an area where power supply is flaky ( upto 8 power outs a day ). EXT2 simply dies and corrupts with fsck. EXT3 looses a lot of transactions and needs a fsck every now and then. reiserfs-3.6 would loose upto 1min+ of work before the power out, unless I mount with sync ( which beats the point of journalling). reiser4 looses only 2~3s of work and never needed a fsck.
- reason 2: It's very effecient at packing files , so conserves space.
- reason 3: I am excited about the idea of plugins ( which are disabled currently, unfortunately ) and want to help finding any bugs quickly, fix them if possible. For example the new xattr API, transparent compression and encryption, etc..
last but not least, stop calling it "ricer", this is provocative and reflective of disrespect. whoever made that name up should be shot in the head. How would you like it if I made fun of your name ?
Anyway, I will have to call it "racer4" then.
But seriously, although the speed issue is controversial, in real life ( desktop ) situations it may be as fast, or almost, as some other filesystems, but not 300% faster like some people claim. As I said I use it for other reasons. These reasons do apply to my desktop and I am willing to loose some responsiveness / interactivity ( if racer4 does actually hurt those, as in severe fragmentation cases ) in order to gain these benifits.
So the argument here is that racer4 is not bad and should not be dissed. In an ideal situation, users should be allowed to choose if they want to try it or not, that's the freedom of open source. Seeing people use it without understanding it, happens everywhere. Hell if love-sources was just tiny patch to add a different scheduler for a certain very specific situation, someone will just use it and pop up the stupid question "what does it do ?" "It's broken" etc.
