[rrd-users] About logging the REAL valuesn

[Tobias Oetiker]

Today you sent me mail regarding Suggestion:

*> Tobi,
*> 
*> There are quite some people asking to keep the stored values.
*> I know, there is an easy workaround if you alter the time stamps.
*> 
*> However, I can imagine something like this (warning: may still have
*> some rough edges):
*> 
*> 1) define a new data source type, for instance "REAL"
*> 2) disallow AVERAGE on this DST.  It is okay to use MIN,MAX and LAST.
*> 3) When data is shifted into an RRA, there is no ambiguity.  Except
*>    for average (which is not allowed) the (an) original value is kept.
*> 4) This even works when consolidating multiple values into one, when
*>    data is moved to an RRA with steps != 1
*> 
*> It does complicate the program (a bit?) and makes life more easy for
*> the humans.  That's exactly where computers are made for.
*> 
*> Perhaps this is worth a discussion on the developers list?

Oh yes ... here goes ... 

<RANT>

the problem is that people are doing things which are not right
todo .. it makes no sense what so ever to sample a data value and
store it at an arbitrary point in time and then forget about the
time ... you could as well use a random generator to 'create' the
data for you ... or then, the whole exercise is not about monitoring
but about inventing statistics ... this is also an interesting
exercise but I would rather not take part in it ... What ever way
you slice or dice it, data samples have a sampling time associated
with them. If you modify that time you are falsifying your data. 

Using MAX and MIN is a different issue all together because rrdtool
clearly defines that it FIRST builds an average value over the
defined step interval and then takes the MIN and MAX of these
average intervals ... If you need more precision (or higher peaks
for that matter) you must have shorter sampling intervals ... in
the case of a counter monitor you could drive this to the extreme
where at an very small interval you will always get a MAX of 100%
... it is a binary world after all ...

In MRTG i made the serious mistake of taking the MIN and MAX values
directly from the real data intervals without pre averaging ...
this is why there are these disturbing posts of people who wonder
every now and then how it can be that the 5 minute MAX values in
mrtg log files are higher than the 5 MINUTE average values ...

To wrap things up: If someone wants to mess with their data and
randomize it at will then they have todo this actively by ALTERING
the time stamp before feeding the data into rrdtool I do not want to
help committing this .... 

</RANT>

OK, this should maybe help to get some discussing started, or what
do you think ?

tobi


-- 
 ______    __   _
/_  __/_  / /  (_) Oetiker, Timelord & SysMgr @ EE-Dept ETH-Zurich
 / // _ \/ _ \/ / TEL: +41(0)1-6325286  FAX:...1517  ICQ: 10419518 
/_/ \.__/_.__/_/ oetiker at ee.ethz.ch http://ee-staff.ethz.ch/~oetiker



--
Unsubscribe mailto:rrd-users-request at list.ee.ethz.ch?subject=unsubscribe
Help        mailto:rrd-users-request at list.ee.ethz.ch?subject=help
Archive     http://www.ee.ethz.ch/~slist/rrd-users
WebAdmin    http://www.ee.ethz.ch/~slist/lsg2.cgi