Re: [pmacct-discussion] reloading config & accuracy

[Tony]

Hi Paolo,

--- On Tue, 22/9/09, Tony <td_mi...@yahoo.com> wrote:

--snip all old discussion--

> 
> Now that adjb seems to be doing what it is supposed to do I
> will accumulate a few more days/weeks of data and compare
> the values from pmacct (with adjb) to those being recorded
> directly by the packeteer. Hopefully they will be a lot
> closer now..
> 

I've accumulated a few weeks worth of data and the results are indeed much 
better.

The variance between the data recorded by the packeteer and pmacct (with adjb) 
for various size subnets is:

1.041%
-0.947%
0.650%
0.466%
0.440%
-8.822%
3.239%
4.202%
-26.559%
-0..042%
-0.979%
-0.041%
0.340%
-0.684%
0.0617%
-1.488%
-0.024%
-0.920%
-35.888%
-0.582%


The reason there are a lot of different individual values is that the packeeter 
is configured to record data into the subnets (ie. /27 - /30) allocated to 
individual customers. I'm then comparing that to the data recorded by pmacct 
for the same IP address range (as extracted from SQL DB). This is the easiest 
way for me to compare without messing with the packeteer.

The aggregate volume of data for the above comparisons is in the range of 500GB 
so is reasonably large enough to give valid statistical results (IMHO).

There are a couple of large discrepancies (26% & 35%), but they are few in 
number compared to the majority of results which are under 1% difference.

My next step is to start gathering the netflow information directly from the 
router instead of getting the packeteer to export it. For those who missed the 
original discussion, I'm currently getting the packeteer to export netflow 
information to pmacct. I wanted to make sure that the netflow information 
exported tallied up reasonably closely to the data being recorded natively by 
the packeteer which they should because the source of the data is identical 
(both from the packeteer).

On the side question I asked about sorting results by IP address, I ended up 
just dumping the values into an XLS spreadsheet, splitting the IP address on 
octet fields (using "." as a seperator for split) and then ordering by the 
third & fourth fields resulting fromthe split. Thanks for the suggestions to 
move to PostgreSQL.


regards,
Tony.


      
__________________________________________________________________________________
Get more done like never before with Yahoo!7 Mail.
Learn more: http://au.overview.mail.yahoo.com/


_______________________________________________
pmacct-discussion mailing list
http://www.pmacct.net/#mailinglists