[DGD] DGD/MP & OpenMosix revisited?

[Felix A. Croes]

Greg Lewis <glewis at eyesbeyond.com> wrote:

> > This point I won't cede.  Some things just don't scale well using
> > clusters, and in my opinion that includes MUDs.
>
> I think thats the interesting question here :).  How hard is it to design
> a MUD which will run in a clustered environment?  I'm interested that you
> see this as quite a lot harder than designing a MUD for an MP system.
> Maybe you could elaborate on that?

That's the whole point of DGD/MP.  It simulates a single-threaded
system using concurrent threads running on multiple processors.
Concurrency is not visible to the LPC programmer.

The only guideline for the LPC programmer is to break up threads that
access a lot of objects, using callouts.  And if he doesn't, all he
loses is efficiency.  There are no mysterious failures or
inconsistencies.

Perhaps you meant to include the difficulty of creating DGD/MP in the
equation?  All I can say here is, while writing DGD/MP, a distributed
system looks a lot harder to me.  I see no convincing designs anywhere,
only the usual zones, toy solutions and handwaving about object
migration this, OpenMosix that, peer-to-peer will-solve-all.

I don't claim that DGD/MP's concurrency model is the fastest possible
one, just that it is a good combination of speed and usability.  You
could transplant it to a distributed network, but in the end you are
going to saturate the bandwidth sooner and you have to cope with
additional timing problems.

I am going to withdraw from this debate for now, until I have a version
of DGD/MP ready to be released.

Regards,
Dworkin
_________________________________________________________________
List config page:  http://list.imaginary.com/mailman/listinfo/dgd