On Thu, Mar 29, 2001 at 10:55:21AM +0200, Alex van den Bogaerdt wrote:
> George Dau wrote:
> > Just fake the timestamp so it looks like it runs exactly on the 5 minute
> > boudary. Instead of using:
> 
> Process the following events using that method:
> (don't say this is unrealistic, it is just an exaggerated example)
> 
> 12:00:00  there are 20 lines in use
> 12:00:10  you monitor
> 12:00:30  incoming call
> 12:01:00  incoming call
> 12:01:30  incoming call
> 12:02:00  incoming call
> 12:02:30  incoming call
> 12:03:00  incoming call
> 12:03:30  incoming call
> 12:04:00  incoming call
> 12:04:30  incoming call
> 12:05:00  10 lines are disconnected
> 12:05:10  you monitor
> 
> What numbers do you get?  20 at 12:00:10 and 19 at 12:05:10.
> You enter this as 20 at 12:00 and 19 at 12:05.
> 
> What is the real maximum? 29, not 20.

  However, this won't be affected in any way by quantizing the time
stamps.  With RRD's usual rules, for the 12:00-12:05 interval the
maximum would be entered as an average of 20 and 19:

  ((10sec*20)+(590sec*19))/600sec, or something like that.

  In either case, the precision is quite high, but the accuracy (as a
reflection of the real events) is very poor.  I don't see that
quantizing the time intervals (to force quantization of the reported
values) hurts in this case. 

  -- Clifton

-- 
 Clifton Royston  --  LavaNet Systems Architect --  [EMAIL PROTECTED]
   WWJD?   "JWRTFM!" - Scott Dorsey (kludge)   "JWG" - Eddie Aikau

--
Unsubscribe mailto:[EMAIL PROTECTED]
Help        mailto:[EMAIL PROTECTED]
Archive     http://www.ee.ethz.ch/~slist/rrd-users
WebAdmin    http://www.ee.ethz.ch/~slist/lsg2.cgi

Reply via email to