On Wednesday, May 16, 2001 10:44:46 AM -0500 Steve Lord <lord@xxxxxxx>
wrote:
>>
>> I hope everyone on the reiserfs lists realizes how informal these
>> benchmarks are. If you want a real idea of how <foofs> performs, you've
>> got to talk with someone who knows a lot about foofs to make sure things
>> are fair.
>
> Thanks Chris, I totally agree, and at the moment I am much too busy to
> spend time in a performance competition. I should also state that I have
> never regarded the different filesystem on Linux as being in direct
> competition with each other, there will always be benefits to using each
> different filesystem for their strong points. Plus having several
> filesystems under active development means that there will be a tendency
> for the developers to make theirs the best, the implementations improve,
> and everyone wins.
Grin, Steve and I have the benefit of having had this conversation before.
Just so everyone on the list(s) knows....we really do agree on this. The
resulting changes to the linux kernel from all these filesystems should
eventually make linux a very strong system for i/o heavy applications.
Plus, the ability to compare these different filesystems under the same
kernel is just plain cool.
>> reiserfs acls are so fast, its like they aren't even there ;-)
>
> XFS ones will be fast too - if they fit in the inode, which can be
> achieved if you know what you are doing (inode size is a mkfs option).
Really, the reiserfs acls aren't even there. Anyone doing benchmarks
should think about the XFS features reiserfs doesn't have, turn some off
(acls), and leave some on (delayed allocation, O_DIRECT).
-chris
|