[hashcash] Re: parallel hashcash proof of concept

>>>>> "Hubert" == Hubert Chan <hubert@xxxxxxxxx> writes:

[...]

Hubert> Running some informal tests, generating a 25-bit stamp in
Hubert> parallel *can* take about the same time as generating a 22-bit
Hubert> stamp with just one process (as would be expected), but
Hubert> sometimes generating the 22-bit stamp is much faster.  But my
Hubert> number of runs is far too small so far to say anything
Hubert> meaningful.  I suspect that the variance in generation time
Hubert> should decrease too.  (It's been a while since I took stats.)

OK, here's the results after 30 runs of each:

22-bit stamp, single process:
       avg.:  31.433s (wall clock time)
  std. dev.:  38.349
       max.: 166.706s
       min.:   0.314s
(huge difference between max and min times)

25-bit stamp, 8 parallel processes:
       avg.: 31.264s
  std. dev.: 20.161
       max.: 75.205s
       min.:  9.029s

So on a cluster machine, it seems to perform as expected.

-- 
Hubert Chan <hubert@xxxxxxxxx> - http://www.uhoreg.ca/
PGP/GnuPG key: 1024D/124B61FA
Fingerprint: 96C5 012F 5F74 A5F7 1FF7  5291 AF29 C719 124B 61FA
Key available at wwwkeys.pgp.net.   Encrypted e-mail preferred.

Other related posts: