[rc5] Anybody notice the abberations in the total proxy rate?

Sanford Olson sanford at msn.fullfeed.com
Thu Jul 31 02:38:00 EDT 1997


[SNIP]
>So, on the list you've grabbed we have proxies that have been up for as
>long as a week and some which have been up only a matter of hours.  This is
>where the descrepancy comes from.  The final entry, which is labelled as 
>"---OVERALL SERVER RATE---" is the calculated rate from the master server, 
>and is probably the best indicator of our true speed.  However, if the
>server has been up for an unusually long time or an unusually short time,
>this figure can be a bit "off".
>
>A more accurate formula could be devised, but at a cost of more cpu cycles
>and more complexity.


How about a moving average?  Slightly more complex, but minimal CPU cycles...
On each proxy server: Allocate a 24 slot array; Once an hour, record "total
blocks received" in the array, if the array is full, lose the oldest entry.
 Then, of the remaining entries, subtract the oldest from the newest and
divide by (24 - 1).  Store this value as the "current average keyrate" for
this server.  Doing these calculations once an hour would minimize the
performance impact.  A more "lively" average would be 12-slots and once
every 5 minutes.

Just a thought...

- Sanford

----
To unsubscribe, send email to majordomo at llamas.net with 'unsubscribe rc5' in the body.



More information about the rc5 mailing list