From: J. Andrew Rogers (andrew@ceruleansystems.com)
Date: Fri Sep 09 2005 - 02:18:26 MDT
On 9/8/05 3:32 PM, "Ben Goertzel" <ben@goertzel.org> wrote:
> Distributed systems, as commonly used in large-scale infrastructure today,
> are *very* different from massively parallel MIMD systems like the
> Connection Machine, in terms of algorithm design and analysis....
This not really true, as I actually work on massively parallel
message-passing systems -- MIMD in short -- for infrastructure systems
purposes where single applications run in bits and pieces on different
machines. There has been a slow trend toward such architectures, though
admittedly we've been on the front edge of that trend. The CM-5 was cool;
I'd have loved to have worked one back in the day.
However:
Off-the-shelf distributed systems that you can put together for modest cash
today will spank the bejeezus out of a Thinking Machines CM-5 in terms of
bandwidth, latency, and of course raw crunch. Moore's law and all that.
You can replicate the model in your basement for not much money, given a
good reason to.
Obviously, people designing MP-based MIMD infrastructure systems will be
solving the same kinds of problems you were solving on a Connection Machine,
particularly since the MP fabric is actually more capable than those old
machines.
I'm probably too close to it such that I think everyone else is familiar
with this. Infrastructure parallelism today is far more advanced than
simply partitioning the application stack vertically, though that is the
type that most people are familiar with. The driving force behind the move
to this type of architecture is latency; geographic distribution and scaling
of single applications that cannot be pooled or replicated due to
consistency/synchronicity requirements is a big growth area.
J. Andrew Rogers
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:52 MDT