Java multi CPU capabilities

[Lennart Sorensen]

On Thu, Aug 29, 2013 at 12:46:48AM -0400, Christopher Browne wrote:
> The fact that the common JVMs are expected to behave properly on
> Certain Operating Systems that don't much like forking processes puts
> a certain amount of "crimp" in this, although I don't see people
> complaining too much that you need to use fork() in order to get the
> average C application to use multiple CPUs...

Of course we don't complain, because we don't use fork.  We use
pthread_create of course.  And even windows that doesn't have fork,
does support creating threads.

> In practice, I'm not sure that many users terribly much *care* if a
> single Java application harnesses all the CPUs on their systems; that
> tends to get covered nicely by such additional factors as:
> 
> a) Composing a broader service using multiple Java applications;
> b) Making the system easier to scale up by having multiple instances
> of interesting Java applications
> 
> As soon as you do those things, you spatter the system with a bunch of
> processes, and harness as many CPUs as you want to.
> 
> It's a pretty common strategy to attain parallelism by partitioning
> problems so that you can throw it against multiple physical servers;
> that strategy works perfectly well at splitting across CPUs.
> 
> If you choose a supremely monolithic implementation strategy for your
> system that makes that nearly impossible, I'm not sure I'd consider
> that Gosling's problem.

-- 
Len Sorensen
--
The Toronto Linux Users Group.      Meetings: http://gtalug.org/
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://gtalug.org/wiki/Mailing_lists