Re: Forcing 2^31 keys/block

Evan Jeffrey (ejeffrey@eliot213.wuh.wustl.edu)
Mon, 16 Jun 1997 18:09:25 -0500


>On Mon, 16 Jun 1997 14:32:13 -0500 (CDT), Timur Tabi wrote:
>
>At the start of the challenge, the DESCHAL client (or the server)
>increased key block size until the computation time would be as
>close to one hour but not more. Some weeks ago, this restriction
>was changed to 2 hours.
>
>[ Now, why you aren't getting 2^31 keys I don't understand, since
>you say it's taking 45 minutes per block on your machine. ]

In addition to the 2 hour target, something is hard coded not to take more
than 2^30 key pairs at once, except for the bitslice clients (changed when
some fast alphas were doing 2^30 pairs in under 15 minutes...)

One solution to the problem would be to, as stated, run two copies of the
program simultaneously. This would allow you to push the limit into the 2
hour range. To go further, you would need to start three clients, force
them to get up to 2^30 each, then let them run. Unfortunatly, at midnight,
they will recallibrate, and you would be back down into the 2 hour range.
Now the trick to getting them sync'ed properly is to disable autodial and
set a cron job to connect at set intervals which are slightly greater than
the time necissary to finish both blocks. The clients will finish, wait on
the network coming back up, and report/start new blocks simultaneously.

You are still going to have problems getting it to stay over 2 hours through
midnight without some really tricky cron stuff, which is going to be
especially tricky on a non UNIX system. You would need to play around with
suspending and restarting processes to get them all going on 2^30 pair key
blocks. I wouldn't recommend it.

===
Evan Jeffrey
erjeffre@artsci.wustl.edu

Let us go. Let us leave this festering hell hole. Let us think the
unthinkable, let us do the undoable. Let us prepare to grapple with the
ineffable itself, and see if we may not eff it after all.
--Dirk Gently