why doesn't this impress me? i'll tell you why. with o a stock version of des (dennis ferguson's), which is written in c, and not optimized for any particular chip or vector hardware o a no-name 50 Mhz 486, which you can buy for under $1,000 at fry's o netbsd, a freely available general purpose operating system i have measured 29,000 des crypts per second. now give me a "1,024 node" machine made of of these -- admittedly unwieldy, but no doubt a hell of a lot cheaper than a 1,024 node CM/5 (and a hell of a lot more useful, imho) -- and i can run at three times the "world record" rate. peter, inveterate iconoclast
peter honeyman says:
why doesn't this impress me? i'll tell you why. with
o a stock version of des (dennis ferguson's), which is written in c, and not optimized for any particular chip or vector hardware
o a no-name 50 Mhz 486, which you can buy for under $1,000 at fry's
o netbsd, a freely available general purpose operating system
i have measured 29,000 des crypts per second.
now give me a "1,024 node" machine made of of these -- admittedly unwieldy, but no doubt a hell of a lot cheaper than a 1,024 node CM/5 (and a hell of a lot more useful, imho) -- and i can run at three times the "world record" rate.
You can do even better if you happen to have 2000 sparcstations which are idle for 16 hours a day. Myself, I wonder how many machines we would need for a net parallel DES crack. Perry
why doesn't this impress me? i'll tell you why. with [use a bunch of PCs running some freenix to do it]
now give me a "1,024 node" machine made of of these -- admittedly unwieldy, but no doubt a hell of a lot cheaper than a 1,024 node CM/5 (and a hell of a lot more useful, imho) -- and i can run at three times the "world record" rate.
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. Either way, I could think of more fun things to do with those 1024 PCs :) jim
Perry Metzger, pmetzger@lehman.com, writes:
You can do even better if you happen to have 2000 sparcstations which are idle for 16 hours a day. Myself, I wonder how many machines we would need for a net parallel DES crack.
I think this is Perry's way of telling us how Lehman Brothers (or Shearson-Lehman, or American Can, or Primerica, or Kuhn-Loeb, etc.) is _really_ making its money! Put those unused CPU cycles to use! -Tim May -- .......................................................................... Timothy C. May | Crypto Anarchy: encryption, digital money, tcmay@netcom.com | anonymous networks, digital pseudonyms, zero 408-688-5409 | knowledge, reputations, information markets, W.A.S.T.E.: Aptos, CA | black markets, collapse of governments. Higher Power: 2^756839 | Public Key: PGP and MailSafe available. Note: I put time and money into writing this posting. I hope you enjoy it.
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?
You can do even better if you happen to have 2000 sparcstations which are idle for 16 hours a day. Myself, I wonder how many machines we would need for a net parallel DES crack.
I've got ~1500 workstations... Most are idle at night...
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.
They do not necessarily have to be co-ordinated (or interconnected at
all). You could easily give each one a range of combinations to try, and
wait until one succeeds.
Ben Byer
participants (7)
-
bbyer＠BIX.com -
Eric Blossom -
Jim McCoy -
Perry E. Metzger -
peter honeyman -
tcmay＠netcom.com -
Timothy Newsham