[RC5] Thoughts on Stats & "Block Hoarders"
silby at silby.com
Tue Apr 20 23:32:48 EDT 1999
On Mon, 19 Apr 1999 17:24:05 -0500, Bill Broecker wrote:
>Here's my $.02:
>If I understand things correctly (and probably don't), older clients
>(everything older than 440?) use(d) a LIFO buffer system. In this case, one
>of these "block hoarders" could be sitting on THE finished key, but may not
>turn it in for some lengthy period of time, because they want to see their
>name at the top of the stats.
If the block is finished and they don't want to flush, LIFO or FIFO
buffers make no difference. In fact, the output buffer hasn't changed
in operation - there's no effective difference if it's LIFO or FIFO.
>As well, if they keep buffering new blocks, or acquired a huge qty of blocks
>a long time ago, but don't regularly complete every block in the input
>buffer and flush them, they could be sitting on THE key for who knows how
Yes, that's certainly true. However, since the key's position in the
keyspace is unknown, we should be concerned primarily with processing
the most keys, and not necessarily in which order they're done in.
>Seems to me (again, I could be all wrong here) that we'd like to see every
>participant regularly consume all blocks in the input buffer, and submit
>them on a regular basis. I've been going out of my way to be sure I
>complete all of the blocks in the input buffer on a regular (at least
>weekly) basis. Additionally, if I take a computer out of service, I make it
>finish the input buffer, and flush the output buffer.
In the long run, the concern isn't as much on emptying the in buffer on
a regular basis as much as not losing blocks on a regular basis. In
other words, having 50 old blocks in your buffer file isn't as much of
a problem as dumping 5 blocks a week because you're worried about them
Mike "Silby" Silbersack
silby at silby.com
To unsubscribe, send 'unsubscribe rc5' to majordomo at lists.distributed.net
rc5-digest subscribers replace rc5 with rc5-digest
More information about the rc5