Re: Performance Issues

[David Sterba <dsterba@suse.cz>]

On Mon, Sep 22, 2014 at 12:37:59PM +0000, Holger Hoffstätte wrote:
> Thanks Dave - that confirms everything I (unscientifically ;) observed so
> far, since I also tried to use "find" to warm up (in the hope it would
> cache the relevant metadata blocks), but running with strace showed that
> it does - of course! - not call stat on each inode, and just quickly reads
> the directory entry list (via getdents()).
> 
> This meant that even after a full "find" a subsequent "du" would still be
> slow(er). Both the cold "find" and a cold "du" also *sound* noticeably
> different, in terms of disk head scratching; find is significantly less
> seeky.

The default find still calls some variant of stat quite a lot because it
needs to decide whether to recurse to the directories, but when compared
to 'du -s' it's still not called for all files.

> Interesting that you also mention the readahead. I've run the "du" warmup
> under Brendan Gregg's iosnoop and it shows that most stat()-heavy I/O is
> done in 16k blocks, while ext4 only seems to use 4k.

Most probably, 16k is the size of the metadata blocks, mkfs option
--nodesize or default since progs v3.12.