Sorry for the late reply, I was busy rebuilding our test lab and had to
leave my Netlfow experiments aside for a while.
Thanks for the your response, I understand the constraints. I will use perf
counters as an aggregator for now, it will probably add some I/O load and
storage needs but all-in-all it is still better than non-aggregating
solutions and - as you said - I can do the aggregation myself.
On Sat, Dec 17, 2016 at 4:45 PM Paolo Lucente <pa...@pmacct.net> wrote:
> Hi Yann,
> You remember i was saying of the current limitations of the
> aggregate_primitives framework. That's it: you can add key primitives
> to the aggregation method but you can't add non-key ones on which, for
> exxample, you want to perform operations (ie. sum like in the case of
> bytes and packets). It is on my todo list to add such feature although
> priority is not high. This all said, what i must be realistic is i was
> not thinking to include 'advanced' operators like average (as that would
> introduce a whole new behaviour, ie. cache entries de-aggregated in
> memory then consolidte on purge [to make the average work]). What i may
> suggest as workaround is to use the aggregate_primitives framework you
> have today and make the averages yourself by post-processing the output.
> On Thu, Dec 15, 2016 at 01:44:16PM +0100, Yann Belin wrote:
> > Hello,
> > I am looking for a way to report on additional primitives with nfacct.
> > It does by default with bytes and packets but I cannot find how to add
> > extra fields. Am I missing something?
> > What I want to include in reports is application performance
> > information (rtd, packet-loss, etc.) from Cisco devices, as described
> > on the link below. For a given flow, I would like to store an
> > aggregated version (e.g. average) of this data.
> > Thanks,
> > Yann
> > _______________________________________________
> > pmacct-discussion mailing list
> > http://www.pmacct.net/#mailinglists
> pmacct-discussion mailing list
pmacct-discussion mailing list