-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 25.02.2014 00:12, Bobby Evans wrote:
> Then there are the latency metrics where it gets even more complex
> because to aggregate them you also need the corresponding event
> counts.  The metrics for these that you get from the UI/Nimbus
> handle this for you, but with this system you need to do some of
> the math yourself to compute a latency weighted by the event
> throughput.

Just as Bobby mentioned there are some metrics (like latency or
rate/s) that you cannot aggregate correctly without having access to
the underlying "raw" counts that make up the metric.  For that reason
the current implementation of Storm's built-in metrics feature has
been only of limited use for us unfortunately.  Instead we have been
using Coda Hale's Metrics library [1] directly and then push its
metrics into a tool like Graphite.  See [2] for a concrete example.

Best,
Michael




[1] http://metrics.codahale.com/
[2]
http://www.michael-noll.com/blog/2013/11/06/sending-metrics-from-storm-to-graphite/
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.22 (MingW32)
Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/

iEYEARECAAYFAlMTexgACgkQeW5XuG18ujTbiACgkj6xYCPAuPzMmGKAUCvMvSPM
020An3IWcdufVBZFLCmOLy2COSDQKYOj
=/G4I
-----END PGP SIGNATURE-----

Reply via email to