Matt Welsh wrote,
> My personal feeling is that there is a lot that will need to
> happen at both the Java and the O/S level to get great I/O
> performance. I am not sure I agree with many of the
> discussions in the linux-kernel list that the right way to
> get high I/O bandwidth is just to use some bastardization of
> signals; I think that the folks at Rice are a lot closer to
> the mark with their novel event-delivery mechanism
Agreed. One big headache is the fact that asynchronous IO
(POSIX or otherwise) is typically going to require that buffers
be at fixed addresses, which is hard to map onto current JVMs
without an unacceptable amount of copying. Event driven
mechanisms (like the one described in the paper you reference)
and Solaris /dev/poll combined with non-blocking IO seem to
be a way of avoiding that problem.
I guess that in principle it ought to be possible to tweak
JVMs to special-case a priviledged class of byte[]s to allow
them to be pinned for an extended interval without completely
screwing GC, but I'm not at all clear at the mo' how that could
be done. By the look of it the latest rev. of Jaguar might
give me some answers.
> You also might want to read a paper we wrote recently about
> engineering systems for high throughput.
Looks very interesting from a quick scan. It looks like we're
going in pretty much the same direction.
Cheers,
Miles
Ps. Looking forward to working with you (Matt) and Dan on
JSR 51.
--
Miles Sabin Cromwell Media
Internet Systems Architect 5/6 Glenthorne Mews
+44 (0)20 8817 4030 London, W6 0LJ, England
[EMAIL PROTECTED] http://www.cromwellmedia.com/
----------------------------------------------------------------------
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
