[RC5] When RC5 will be finished...
tomdv at datatx.com
tomdv at datatx.com
Sun Nov 4 12:41:50 EST 2001
Dear Bruce and others,
I have read all your comments with interest. But still, my old P120 goes
off line. The client am working with, allows for 'random' keys in the RC5
contest. So, this random key goes against the 90 days rule, mentioned
earlier. If a computer takes too few blocks, or work units, it runs empty
so it starts with 'random', just happening to take the blocks of my good
old P120 away. But now that it has only a 3 day recycle period, that
definitely rules out the P120.
See, my point is that apart from improving the client, we can also think
about the word 'distributed', which is where also a lot off waste can be
generated. As for knowing where what went, we all have a participation
number, right? That might be enough, but that it will be a hard task to
do, I can see that from where I'm sitting.
Did you do any calculations on what the effect might be of the 3 day
recycle, with random blocks sitting in the way as well, with older
computers that run slower, and so on? Would it be useful to do it?
By for now,
Please respond to rc5 at lists.distributed.net
Sent by: owner-rc5 at lists.distributed.net
To: <rc5 at lists.distributed.net>
Subject: RE: [RC5] When RC5 will be finished...
>I believe this question also goes for RC5. We had this issue before
when we were speaking about load of work on a PC. See, if I use a P120
and only update every two weeks, I have a pretty high chance the blocks
it crunched have been processed already by someone else. Actually, I
deinstalled dnet from that computer for exactly that reason.
This is not true. We avoid recycling work units for several months to
avoid this exact problem. If you download umpty-jillion work units, you
could run into this on any speed PC. If you stick closer to our
recommendations, you're not vulnerable.
>For all at dnet, we always speak about 'wasting' cycles when we
mention a screensaver. Now, apart from that, if we have slow computers
that update every two weeks only and their work was done by a faster
computer, that also is a waste of cycles. Rather than starting any new
contest, is it possible to review what can be done to avoid this from
happening or at least have a better control? Maybe you need to have a
list of who has what units and verify what happen with the work units if
long time no see.
We are already struggling with a huge data load just managing the stubs.
Trying to track where we sent the work (which proxy did it go to? Did
it go to a perproxy or a client? Was the work unit processed by that
client, or did a different computer do it via sneakernet or shared
buffers? How could we uniquely identify everyone who checks in for
work?) would be a huge and impractical undertaking.
Bruce Wilson <bwilson at distributed.net>
PGP KeyID: 5430B995, http://www.toomuchblue.com/
rm -rf /bin/laden && chmod 0777 /dev/freedom
To unsubscribe, send 'unsubscribe rc5' to majordomo at lists.distributed.net
rc5-digest subscribers replace rc5 with rc5-digest
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the rc5