Nelson Minar wrote:
> What a great discussion this has been!
>
> >I don't think anyone is arguing that high thread counts are always
> >wrong or that the current implementation is the best of all worlds.
>
> Phew!
>
> >Yes, the Linux kernel can do better and Java implementations can do better
>
> And in the current situation, we have control over the Linux kernel
Careful... there was some more context around my comments. There's a reason
the upcoming async I/O additions are called the "Scalable I/O API", and are
showing up due to heavy demand from the Java community. Even Solaris, with its
boasts of a highly tuned threading system, hasn't been such a happy place to
build highly scalable Java server products. Yes, Linux can do better, but it's
still a very long way to the "threads are free" ethos behind the current,
flawed Java I/O model.
Nathan
>
> (it is open source, afterall) but small access to the Java
> implementations. This whole thread started when I said that one of the
> Sun engineers claims that Hotspot can't really work with green
> threads. That means that Sun's going to rely on native threads only.
> The only option left is to bang Linux' native threads into shape.
>
> Does anyone have good numbers on how well Linux threads scale? I don't
> really know, I'm only talking from the vague understanding I have that
> Linux native threads aren't as good as Solaris or NT.
>
> [EMAIL PROTECTED]
> . . . . . . . . http://www.media.mit.edu/~nelson/
>
> Make your computer useful 24 hours a day: http://www.popularpower.com/
----------------------------------------------------------------------
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]