[RC5] farfetched stats idea?

Derek Berquist dberquist at marmen.com
Tue Jun 27 18:32:35 EDT 2000


I think that there are two major items to consider.

1) Is there any possible benefit to dnet?  A person would either be
turned off by being declared "obsolete" and possibly quit.  Or a person
could be proud of being in the obsolete class, thus encouraging them
to surely not start any other computers on the task lest they lose 
their "obsolete" rating.  Neither has a positive effect for dnet.

2) Would it have any relation to reality?  You may be right that *most* 
people use a single machine.  However, many use significantly more than
one.  Where does that bring the average?   1.3?  1.5?  2.0?  This is
difficult to judge.  Add to that the fact that many do not run 24 hours
a day.  What percentage does the average run?  Again difficult to say.
Now combine those two factors and try to get a meaningful number.

Here are three simple scenarios that would result in unfair obsolete
classification.  1)  A person runs an array of 5 or 6 386's.  The total
computation results in a high enough number to avoid obsolete rating.
However few people would argue that the 386 was not obsolete for general
business use.  2)  A person has a blazing 800 MHz (or whatever) machine
at home that is used an average of 1 hour a day on  weeknights or 3% usage
overall.  This machine would be seriously in danger of being declared
"obsolete."  3)  A middle of the road machine that actually gets high
utilization as a server/gateway/remote access/whatever.  Even though the
processor is relatively current, the dnet performance is lacking, and thus
considered obsolete.

Some of the technical issues could be overcome by referencing processor
information returned with the packet.  (I *think* that info is there, 
but I don't know firsthand.)  Then somebody could just "declare" a 
certain processor obsolete.  However, in addition to being subject to
whimsy, this would not overcome issue #1 above.

I will agree with your comments on one point.  It is indeed, "a queer 
idea for stats."

Derek

> Hi Decibel et al. I got a queer idea for stats:
> A timeline: "When did your CPU become obsolete?"
> 
> Last I checked Dnet does 147.9 Bk/s, with 44,362 people active.
> Assuming *most* people use a single machine to conribute keys, that
> comes out to ~3.3 Mk/s. Now let's pick an arbitrary threshold--5%--and
> say if your processor at best can't do more than that (0.05 x 3.3M = 165
> kk/s, which I think is about as fast as a 486-133?), then it is
> considered "obsolete". 
> 
> So you look at the historical performance du jour of Dnet, and average
> peaks for processors we have stats for, and the first time Dnet's
> average user's contrib is over 20x (if we use the 5% threshold) your
> processor, then it gets a seat in the timeline, lucky you!  :)
> 
> --GAck

--
To unsubscribe, send 'unsubscribe rc5' to majordomo at lists.distributed.net
rc5-digest subscribers replace rc5 with rc5-digest



More information about the rc5 mailing list