[RC5] Happy Holidays from distributed.net
nugget at distributed.net
Mon Dec 14 06:00:17 EST 1998
Happy Holidays from distributed.net! It's been nearly a month and a
half since the last official "state of distributed.net" update and
we've got quite a lot of news to cover this time. The success of
the online .plans (http://nodezero.distributed.net/cgi/dnet-finger.cgi)
has taken some of the pressure off, it's still past time for an
Most notably, we broke 4% on 6-Dec-1998. It took us 48 days to go from
3% to 4%, versus 61, 92, and 211 days to go from 2% to 3%, 1% to 2%,
and from 0% to 1%. For the week ending 12-Dec-1998, we did 110.8
million blocks for an average keyrate of 49.16 Gk/s. This is an
increase of 466,979 million blocks and 1.45 Gk/s from the prior week.
At 49.16 Gk/s, it will take 11.43 years to exhaust the keyspace, down
from 11.8 years last week, and 78.96 years after our first month on the
project (22-Nov-1997). If our rate is growing linearly at 65,550
blocks per day, we will exhaust the keyspace in 3.3 years. If our
keyrate is doubling every 156 days, we will exhaust the keyspace in 1.8
years. There are currently 55,866 emails that have submitted blocks in
the last 30 days (out of a total of 118,510 emails). At 46.2 Gk/s
RC5-64, our theoretical maximum DES keyrate is 151.9 Gk/s, but this
number could be off by 10% or more.
On the immediate horizon we've got the upcoming DES-II-3 contest in
mid-January which has attracted most of our energies. Thanks to
our colleagues at EFF, the bar has been raised considerably higher for
this contest but we have every intention of facing the challenge.
We're over twice the size we were last time and new client enhancements
from the coding team will go a long way towards a smoother and more
timely commencement. If you think we're not up to the task, you
simply haven't been looking lately.
Realizing that a non-trivial number of our participants were not aware
of distributed.net last July, let me set the stage for this very
exciting undertaking. DES-II-3 is the third of an unspecified number
of DES challenges sponsored by the folks at RSA Labs. It differs from
the present RC5-64 challenge in that this challenge has a time limit.
There are three tiers for success in any DES-II challenge, determined
by how rapidly the previous challenge was successfully completed. With
RC5-64 there is no stated deadline for success but with DES-II-3 we must
recover the correct key and submit it to RSA Labs in under 14 hours to
collect the entire US$10,000 prize. A 28-hour success yields US$5,000,
after which the prize drops to only US$1,000. The contest ends at the
42 hour mark, after which no prize is available.
While US$10,000 sure is a lot nicer than US$1,000 the more important
consideration is that these DES challenges force us to excel and have
proven to be excellent motivator to improve, recruit, grow, and
surpass our previous goals. There's nothing like a ticking clock to
help one focus on the task at hand, and DES-II does this well.
A great number of interesting changes and improvements have been
quietly taking shape over the past months, both on the client side
and the server side.
The following is a brief list from Cyrus Patel (cyp at distributed.net) on
the more significant client-side changes that are nearing final form:
a) In order to allow keyservers to transfer blocks more
efficiently, network block transfers (fetch/flush) are
in the process of being optimized. The client is already
capable of flushing up to 8 times faster (to the existing
proxy network) than it could before.
b) The client's limit on the maximum block size of 2^31 will
be raised to 2^32 in time for DES-II-3.
c) The disk buffer code has been rewritten from the ground
up and is now being optimized to allow for more efficient
sharing of blocks between clients, larger block
sizes and a single buffer file format that support both
in- and out- bound blocks for all contests.
d) Checkpointing has been optimized to reduce the number
of disk accesses. Checkpointing now uses only one buffer
file for all contests.
e) The limit on the number of processors has been removed.
f) Threads no longer buffer two blocks. This allows them to
switch contests faster.
g) Work is in progress to remove the DES 2 thread limit for
(non-mmx) x86 clients.
h) All clients are hup-able/restart-able from the command line
which allows users to set up cron jobs to force the client
to switch contests.
i) Although the client already writes changed contest state
information via the configuration file, the current ini
format is too inflexible to allow sharing. This will
change in time for DES-II-3.
Client side changes aside, there's also been work done to the
keymaster/keyserver network to assist us in putting DES into an early
grave. In order to test and tweak the server-side changes, we've
decided to run a DES Preliminary Test Challenge on our own, prior to
the actual DES-II-3 challenge. This voluntary contest will allow us
to tweak the new settings and code in time for a smooth, flawless
run in January.
: DES Test Contest news from Daniel Baker (dbaker at distributed.net) :
In an effort to fully test all of the DES components of our network in
prepreparation for DES-II-3, we plan to run a full DES contest
simulation. All clients with DES enabled will autoswitch to DES during
this time as they would in a normal contest.
Purposes of this test contest:
- Allow us to practice the DES-enabling process on the keymaster so
there is no confusion on the morning of January 13th, the actual
contest's start date
- Test the rapid deployment of DES contest data and blocks across the
- Obtain an accurate network speed estimate, allowing us to predict the
time required to check the keyspace in full during the real contest.
This simulation will begin on Saturday, January 2nd at 09:00 PST and
will run until we find the key. There will be no cash prize associated
with this test run, but your efforts and CPU power will lend a big hand
in helping us prepare for DES-II-3.
: Plan news from Evan Vetere (vetere at distributed.net) :
In response to public demand - well, #distributed demand, actually, on
IRC - I have set up the plans at lists.distributed.net mailing list.
Every day, at 00:00 Universal Time, a mailing is made containing all
the plan updates of the previous day. One a day, and nothing more - if
no updates are made, the mailing will be skipped. If you're tired of
hitting the web server periodically for updates and finding nothing, or
just tired of hitting it in general, send a mail to
majordomo at lists.distributed.net with the body 'subscribe plans'
followed by your email address.
: distributed.net promotional items, aka distributedWear :
: from Adam L. Beberg (beberg at distributed.net) :
As you may already know, we have been talking about having a line of
t-shirts, mouse pads, posters, and other miscellaneous items to promote
distributed.net. Now that our logo is finalized, we are working on
finding a fulfillment house that can ship items to the many people who
have expressed an interest in such merchandise. Given our international
userbase (over 100 countries), this is a non-trivial task and we have
yet to find an appropriate firm. We also need your help in deciding
which items we should focus on. A survey page where you can make your
voice heard is at http://www.distributed.net/cgi/survey.cgi
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Size: 366 bytes
Desc: not available
Url : http://lists.distributed.net/pipermail/rc5/attachments/19981214/f18ce8a0/attachment-0001.bin
More information about the rc5