[RC5] Block size, fetch/flush frequency and database workload

Mike Silbersack silby at execpc.com
Fri Mar 6 11:23:16 EST 1998


On Fri, 6 Mar 1998 15:31:17 +0200, Hannu Nyman wrote:

>Since it seems to take so long to update the statistics database for
>the RC5-64 contest, it seems to me that we have too many
>transactions. Having so many transactions decreases our efficiency.
>
>I see two factors causing this:
>1) People are processing (too) small blocks, which increases the 
>amount of block transfers.
>2) People are updating their buffers too frequently. (in order to 
>advance in statistics?)
>
>The workload of both the keyservers and statistics database would 
>decrease if there were less transactions. So, should we have a 
>recommendation about using larger block sizes? Like "use a block 
>size that checks a block in about 2-3 hours". (Or use even longer 
>taking blocks.)
>
>Using the default buffer size (10 blocks) with "2 hour blocks" would 
>produce one update per day (which would just perfect for maintaining 
>the position in stats ;-).
>
>Hannu Nyman

You're perfectly right about the block size issue, which is why default blocks are 2^30 as opposed to 2^28 now... if 
people went to 2^31 our load would be half as much again.  However, the time and interval that blocks are sent in is 
irrelevant; it's actually better to send them in as you feel like it so this it smooths out load over the day.

Mike "Silby" Silbersack
silby at execpc.com
Distributed.Net Tech Support

-------------------------------------------------------------------
We cracked RC5-56.  We cracked DES.  Can you help us with RC5-64?
Put your idle CPU to work:  Add it to the world's LARGEST computer.
http://www.distributed.net/
-------------------------------------------------------------------
--
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