Re: Server Keyspace Sizes

Rick Hornsby (hornsby@cis.ohio-state.edu)
Sun, 11 May 1997 17:49:35 -0400


Unfortunatley, it does seem to be causing problems for my win95 486DX4/100
box that the key blocks I've been getting have increased from 2^28 to 2^29

Processor 1 -- Elapsed time: 22229 seconds
Processor 1 -- Key not found
Processor 1 -- 2^29 complementary pairs of
Processor 1 -- Elapsed time: 67226 seconds

It very well could be that I activated the wrong screen saver while I was
gone this weekend and that it chewed up my processor. But a 2^28 still
takes around 40 minutes to complete. I'm not sure that anything can be
done about it at this point, except to wait for the a release that allows
one to choose the max keyblock size :)

btw: Rocke, whoever else wrote this client, y'all have apparenltly done a
pretty darn good job, seriously. I paritioned off some of my HD for linux
and ran deschall on there, and there was very little performance increase
over win95 - which means to me that the client is pretty well written to
take advantage of whatever OS its running under - including 95. Keep up
the good work guys :)

-Rick

At 16.51 05.10.97 -0600, DES Challenge Lists wrote:
>DESCHALL users:
>
>Shortly after midnight (MDT) this morning, I changed the server to send
>larger blocks to the new bitslice clients. [The previous maximum of
>2^30 pairs only took 12 minutes at 3 MKPS. :) ]
>
>I also changed the server to send blocks twice as large as previously.
>(Up to the 2^30 pair maximum for non-bitslice clients.)
>
>Consider this an experiment. If this causes serious problems, please
>let me know.
>
>Yesterday, the raw packet logs were 50MB in size. The postprocessed logs
>were another 18MB. This simple change has lowered the load on the server
>from about 50% to around 30%. And the logs are running about 35% smaller
>than yesterday.
>
>Cheers!
>
>-- Rocke
>
>