From: Timothy Newsham <newsham@wiliki.eng.hawaii.edu>
Date: Wed, 18 Aug 93 17:35:51 PDT
To: mccoy@ccwf.cc.utexas.edu (Jim McCoy)
Subject: Re: World record in password checking
In-Reply-To: <199308182228.AA22350@tramp.cc.utexas.edu>
Message-ID: <9308190031.AA20910@toad.com>
MIME-Version: 1.0
Content-Type: text/plain


> 
> Perhaps because internal communication between those 1024 machines will be
> significantly more difficult than running on a machine that is optimized
> for parallel operations, RPC just doesn't cut it.  You would probably lose
> a number of your hosts off the top just to coordinate the activity of the
> remaining machines. Besides, if you really want to do this spend your
> one or two million (approx cost of your 1000PC site) on seriously dedicated
> DES-cracking parallel hardware.  Do the cracking in hardware, not software.

If you have the pc's aranged nicely there are very few packets that need
be sent.  You can use broadcasts on each net, or multicasting (broadcasting
to a group).  You simply need to send out the password entry to crack.
You could break up the job space by networks, and on each network have
the machines negotiate for portions of the sub-job space.  Alternatively
you could have all machines attacking the key space randomly which
is not as efficient but still quite workable.  Finally if/when one
of the boxes gets a solution, it shouts 'i got it'.  It can broadcast
the solution, which will turn off all the other boxes and get
put onto some consol window somewhere, or some file.

> 
> Either way, I could think of more fun things to do with those 1024 PCs :)
> 
> jim
> 
bon fire?