Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Date: Tue, 16 Jun 2009 09:28:16 -0600
From: RB <aoz.syn@...il.com>
To: john-users@...ts.openwall.com
Subject: MPI John (was: Re: GI John)

On Tue, Jun 16, 2009 at 08:25, Rich Rumble<richrumble@...il.com> wrote:
> I found an article recently detailing how to use MPI and JTR. I hope it may
> help others in their efforts to distribute JTR to many cores and computers.
> http://www.tmto.org/docs/MPIandPasswordCracking.pdf

It's important to note that current MPI implementations of JtR are
only effective in incremental mode.

Curious that he uses Gentoo, but doesn't use the built-in package
manager that not only handles the OpenMPI-1.3.2 install, but also
applies a more recent patchset to JtR with the 'mpi' USE flag.  You
can lead a horse to water...

Is anyone interested in a 9th revision of the MPI patch?  I have it
pretty much done, it just requires a little cleanup before I release
it.  There was zero response on the john-mpi list, so I presumed all
interest was dead.

The mpi9 patch is a descendant of my 'mpi8-small' patch that stripped
out all the extra patches and is specifically designed to cleanly
cooperate with the latest (all-5) jumbo patch.  It has been stripped
down even further and specifically modified to be as non-intrusive to
the original codebase as possible, to the point of even re-writing
portions.  Other than minor compatibility fixes, I don't foresee any
further major updates to this lineage of parallelization.

All told, after all the hype I'm actually a little disappointed in how
little the MPI patches actually do.  When stripped down to brass
tacks, the currently maintained lineage (from Ryan Lim through
bindshell.net) just wraps the JtR main loop in some MPI sugar, then
when in incremental mode splits the cracking space up evenly among the
participating nodes.  No synchronization, no real distribution of the
processing; in fact, it's not terribly different from Solar's current
parallelization suggestion of splitting the workload by namespace
lengths.

So: is there enough interest in the 9th (and probably last major)
revision for me to go clean up my comments and publish it?

-- 
To unsubscribe, e-mail john-users-unsubscribe@...ts.openwall.com and reply
to the automated confirmation request that will be sent to you.

Powered by blists - more mailing lists

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.