Re: *Real* Distributed Computing
- From: AndyW <foo_@xxxxxxxxxxxxxxxx>
- Date: Wed, 10 Aug 2005 21:41:43 +1200
On Wed, 10 Aug 2005 01:15:07 -0400, "Michael N. Christoff"
<mchristoff@xxxxxxxxxxxxxxxxxxxxxx> wrote:
>Hi.
>
>Sorry about the multi-posting, but I wanted to get a general opinion of
>those online. The groups I chose (while on the surface may not seem like
>they would have a lot in common with distributed computing) all intersect
>with the idea of collaboratively solving problems in an environment where
>one may not have a complete set of information. Where there is no 'global'
>entity that sees the global state of the system - ie: where each
>participating entity's view of the world is relative to the other entities
>in direct proximity to it. Anyways, on to the original post:
>
>I was wondering if anyone in this group is looking into the theoretical side
>of distributed computing. I keep reading definitions of DC to the effect of
>"DC is a method of computing where one a) breaks a large problem into
>smaller parts, b) distributes the partial workloads to a set of processes
>that compute the partial solutions, c) finally recombining the partial
>solutions into a total solution". ie: SETI, etc...
>
>This is an *example* of distributed computing - a particular, and very
>straight-forward use of many computers, but is by no means a definition of
>DC.
>
>Does anyone here look into things such as decision tasks (the renaming
>problem, k-set agreement, consensus) or any theoretical papers in the field
>(FLP, the asynchronous computability theorem, the relationship between
>algebraic topology and distributed computing, dihomotopy, ditopological
>homology groups, various calculi for concurrent systems, fine-grained
>concurrency, fault-tolerance, provable impossibility of certain tasks,
>computability in asynchronous, semi-synchronous and synchronous networks,
>worst-case message complexity of distributed tasks, randomized distributed
>algorithms, self-stabilization, shared-memory vs. message passing models,
>fault-tolerance, Byzantine failures, crash failures, omission failures, etc,
>etc, etc...)?
>
>Check out the following link to get a small peek into what I mean:
>
>http://scholar.google.com/scholar?q=algebraic.topology+distributed.computing
>&ie=UTF-8&oe=UTF-8&hl=en&btnG=Search
>
>I'd be interested to know if these things are still entirely a part of
>academia or whether there are people out there who discuss these things in
>their spare time.
>
>
I've spent a decade or so working in this field and its quite a big
one.
Seti is just an example of problem breakdown. Basically they have
many tasks that can be done in parallel and instead of using parallel
computing, they distribute each task onto a different machine. Its a
poor mans example.
The example I like to use is that I have a customer business object
instantiated on a machine in china and a person in the US wants to
update that customer, so his machine pulls the object over and works
on it, then pushes it back. Meanwhile the person in singapore sees
the changes in his application while they are happening. This is an
example of data moving over a hetrogenus network and event
notification propagation. This is a network centric model.
You'll find most common examples use CORBA as that is precisely what
its for. Also, have a look at the way SOM and DSOM worked in OS/2.
DSOM was used for DC.
.
- Prev by Date: Testability and class overhead
- Next by Date: Re: State Pattern: Where to put shared behavior?
- Previous by thread: Testability and class overhead
- Next by thread: Re: *Real* Distributed Computing
- Index(es):
Relevant Pages
|
Loading
