stupid UFS behaviour on random writes

Wojciech Puchar wojtek at wojtek.tensor.gdynia.pl
Fri Jan 18 09:09:16 UTC 2013


>
> But I doubt that such a change would improve performance in the

you doubt but i am sure it would improve it a lot. Just imagine multiple 
VM images on filesystem, running windoze with 4kB cluster size, each writing something.

no matter what is written from within VM it ends up as read followed by 
write, unless blocks are already in buffer cache.



More information about the freebsd-hackers mailing list