>>>>> James Caple writes:
James> In sum:
James> * The Inprise Java-Linux port could not handle nearly as
James> many simultaneous server-side connections as the
James> Blackdown port when using green threads.
And our next release will use poll() instead of select() which will
remove another limitation in the green threads VM.
James> * Using native threads, however, the Inprise port wins
James> hands down. This, of course, is to be expected as the
James> Blackdown porters have not finished testing and fixing
James> their native threads port yet. Using Inprise's native
AFAIK their native threads is more or less a snapshot of our native
threads code. The next release will the fix the remaining problems we
know of: interrupted() problem, interruption of the primordial thread,
EBADF messages.
James> threads, our application was able to handle many more
James> simultaneous connections than we could with the Blackdown
James> green threads port.
[...]
James> One question I have of Blackdown is why they have
James> prioritized their green threads porting efforts over native
James> threads, when native threads offer the best overall
James> server-side performance (if I'm not mistaken in my
James> assumptions)?
We haven't!
The native threads VM is used by default in our JDK. Sun/Inprise uses
green threads VM as default and does not yet support the native
threads VM.
>> As much as I hate to say it, my money is on Imprise. I am finding that
>> Blackdown's VM is causing JRun to hang. This may be a problem with
>> native threads, or with the fact that I am running on a two processor
>> SMP configuration. However, switching to Inmprise / green threads did
>> the trick.
One problem is that there are still several race conditions in the
LinuxThreads implementation. It looks like glibc-2.1.3 will solve a
lot of them.
Juergen
----------------------------------------------------------------------
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
