Re: amr Serious performance issues

[Hans Rosenfeld <hans%NetBSD.org@localhost>]

On Tue, Jan 11, 2011 at 04:46:28PM +0000, Eduardo Horvath wrote:
> On Tue, 11 Jan 2011, Hans Rosenfeld wrote:
> > I don't remember whether I used any partitions at all. The machine is
> > currently down and far away, so I can't check that right now. I think
> > most of the time I tested with dd directly to the raw device, using
> > block sizes 128k and 512k. Is a specially aligned partitioning supposed
> > to make that much of a difference?
> 
> I hope you're not comparing dd performance to the raw device between Linux 
> and anything else.  Linux doesn't actually have any raw devices.  
> Everything goes to the buffer cache, hence any performance numbers you get 
> that way are bogus.

I'm aware of that, but I thought that using oflag=direct with dd would
make the numbers somewhat more useful.


Hans


-- 
%SYSTEM-F-ANARCHISM, The operating system has been overthrown