Getting locks and sharing: was RE: Getting concurrency

[Thomas Lindgren]

--- Vance Shipley <vances@REDACTED> wrote:

> I'm not suggesting anything that complicated just
> that the distribution
> would come with a version of "erl" which would
> detect the presence of
> multiple CPUs and start one node bound to each.  In
> this way there is
> an easy, integrated, supported way of taking
> advantage of SMT.  The two
> main points being that a) the user didn't need to
> think about it and
> b) the nodes are bound to specific CPUs instead of
> left to fend for 
> themselves (not that I know anything about how this
> works).
> 
> I'm not sure what to do after that.  Do you make the
> "extra" nodes
> hidden and try and load share behind the scenes or
> do you just leave
> it at that and expect people to use normal erlang
> distribution?

I think this sounds like a straightforward, simple way
forward. We can start experimenting right now, then
devise ways of simplifying and optimizing systems
consisting of large numbers of (perhaps very closely
connected) Erlang nodes as we gain experience.

This approach seems like a low-hanging fruit to me.

Best,
Thomas



		
__________________________________ 
Discover Yahoo! 
Use Yahoo! to plan a weekend, have fun online and more. Check it out! 
http://discover.yahoo.com/