[RC5] Funky keyrate calculations in proxyper

Cedric Tefft cedric at earthling.net
Thu Jan 22 23:30:45 EST 1998


Jeff Lawson wrote:
> You are probably aware that the proxy does sliding window averages for
> the
> rate calculation.  This requires separating reported blocks into
> discrete
> time chunks and keeping the last x number of chunks for computation of
> the
> rate.  As such, when the oldest time chunks are moved out you will see
> the
> jumps that you have noticed.  For reference, the proxy uses 5 minute
> time
> chunks, and computes the rate from the last 6 of these chunks.

This does explain what's going on, but I have to wonder about the choice
of such a small window.  Even a reasonably fast client that buffers only
five blocks at a time won't check in more often than every hour or so. 
If you are calculating the keyrate based on the blocks returned within
the last 30-minute window you're almost guaranteed to entirely exclude
some of the clients that are served by that proxy at any given time.  In
that case, you will never get a very accurate picture of what your true
keyrate is because the calculation is influenced more by how often a
client checks in than how fast it is.

It seems a better solution would be to forget about sliding windows
altogether and calculate the keyrates based on running totals since the
proxy was last started.  If you do insist on sliding window averages, I
would think a window of many hours (48?) would give a more accurate
average than the 30-minute window currently in use. 

- Cedric

/----------------------------------------------------------\
*      Cedric Tefft: Proto-Human, Reluctant Programmer     *
*            email: cedric at earthling.net                   *
* PGP FingerPrint: E114-CA05-E498-8A88-0762-79BB-A117-C59B *
\----------------------------------------------------------/

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



More information about the rc5 mailing list