[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [DotGNU]DotGNU task list
From: |
Jonathan P Springer |
Subject: |
Re: [DotGNU]DotGNU task list |
Date: |
Sat, 9 Feb 2002 07:55:51 -0500 |
User-agent: |
Mutt/1.3.27i |
On Wed, Feb 06, 2002 at 04:52:51AM -0800, Bill Lance wrote:
>
> --- Sarath B Nair <address@hidden> wrote:
> > --- Bill Lance <mailto:address@hidden> wrote
> > >if you wish to write a customized program module,
> > >then run that remotely.
> >
> > This is what i mean.
> > Ideally i should be able to assign different
> > components of the assembly to
> > different servers(hosting *this* service).
> >
>
> Hummm ... that may be possible in the VRS model, but I
> doubt that it would be practical. The main purpose of
> the VRS is to expose services to a network,
> presumabaly for others to use. Also, if it going to
> actually execute in the VRS, it will have to be able
> to be compiled to an intermediate byte code,
> presumable with Pnet. The VRS incures significant
> overhead to accomplish it's distributed processing and
> storage task.. If this is a purely private function,
> and you have resources similar to those needed for a
> VRS, (for the forseeable future, this means a
> resonably current Linux system) I would think you'de
> be better off with more traditional approachs.
>
I don't want to lose sight of the "address@hidden" distributed computing
possibilities of the architecture. Since much of what DotGNU envisions
requires P2P-like distributed networking, distributed processing
services are a distinct possibility. For example:
1. I want to be rich.
2. I devise some sort of Monte Carlo derivative modelling scheme and
code it in some sort of VRS/DEE aware implementation.
3. Part of that implementation is the function
"myReturnType ReallyComplexAlgorithm(MyInput1 i1, MyInput2 i2)".
4. I call some (hypothetical) service
"distributedComputation(ReallyComplexAlgorithm, distributedInputs[])".
5. The distributed computation starts doling the computations (RPCXML's
containing bytecode and input?) out through a network of P2P-connected
VRS/DEEs. The algorithm for distribution is left as an exercise for
the student :-).
6. The returns come back (perhaps as events?), the core program
processes them and displays them to me, I invest in stock and lose my
shirt.
7. Voila! The cause of good has been served.
Just a (peyote) vision for what this could do to harness all those idle
processors with TCP/IP addresses.
-js
--
-Jonathan P Springer <address@hidden>
------------------------------------------------------------------------------
"A standard is an arbitrary solution to a recurring problem." - Joe Hazen
- Re: [DotGNU]DotGNU task list, (continued)