Buchan Milne wrote:
> No, but this is what Civileme (the highly-respected-in-Mandrake-circles
> QA manager at Mandrakesoft) posted on the Mandrakeforum:
Those figures are grotesquely wrong.
They show ext3 in writeback mode as being slower than in ordered-data
mode, which is unlikely to be the case.
They show ext3 in ordered data mode to be slower than ext2 at reading,
which is not possible - the read code paths are, for all intents and
They show ext3 in writeback mode to be half the read speed of ext3 in
ordered-data mode. But the read code paths are identical.
And the figures appear to show, with no commentary at all, that ext3
deletes files 200x more slowly than ext2, which is simply daft.
All this would tend to cast some doubt over the figures for the other
I'd suggest you put ths one gently aside and run your own tests.