Re : article about More New I/O APIs for Java

[Dupont jean <do...@yahoo.fr>]

It is saying that due to native file systems possibilities the new asynchronous API of JSR 203 will get  best performance  on windows systems  whereas the  non blocking io  API will have give best performance on UNIXes as  they don't provide full robust async OS-level support (at the time of the article publication). 

Article @ http://www.artima.com/articles/io_design_patterns2.html proposes a solution to achieve a platform-independent solution with a common external 
interface using software emulation asynchronous IO API, when OS doesn't natively support it, without performance degradation.

This article clearly states that async API when supported by OS natively gives better performance results. 
But it also seems that efforts since the is areticles has  been published (November 25, 2005) have been made in this way by the Linux community with the AIO library that provides native async support.

So IMHO to achieve the best performance available, it will be necessary to query the system about the available native API. I currently don't know if such possibility is or will be available on the JSR 203 proposal.

my 2 cents :)
Edouard De Oliveira

----- Message d'origine ----
De : James Im <im...@hotmail.com>
À : dev@mina.apache.org
Envoyé le : Mardi, 22 Mai 2007, 8h09mn 55s
Objet : Re: article about More New I/O APIs for Java

In the end, when talking about the asynchronous API versus the
non-blocking API, I didn't get which one will give the best performance.

Is he saying that it depends on the os?

peter royal wrote:
>
>http://www.artima.com/lejava/articles/more_new_io.html
>
>-pete

_________________________________________________________________
Download din yndlingsmusik på MSN Music:  http://www.msn.dk/music  - det er 
nemt og billigt








      _____________________________________________________________________________ 
Ne gardez plus qu'une seule adresse mail ! Copiez vos mails vers Yahoo! Mail