Re: [Aoetools-discuss] [ANNOUNCE] ggaoed 0.9

[Gabor G. <go...@sz...>]

On Fri, Jul 18, 2008 at 01:28:41AM +0200, Torbjörn Pettersson wrote:

> Have you done any performance comparison between ggaoed, qaoed and vblade?

Not really. I've planned to do it but so far I have not managed to put
together a stand-alone test system where the measurements are not
influenced by other services using the box.

> Im interested in the performance differance between aio and normal
> blocking io.

What do you mean by "performance"? Current Linux AIO still has issues
with buffered I/O, so if you only care about streaming throughput,
then the good old read()/write() may still be the best.

OTOH AIO + direct I/O seem to provide nice low latencies, so if you want
to issue a lot of random I/Os to a lot of independent devices then AIO
may be better.

Gabor

-- 
     ---------------------------------------------------------
     MTA SZTAKI Computer and Automation Research Institute
                Hungarian Academy of Sciences
     ---------------------------------------------------------