[j-nsp] Accuracy of interface stats

Richard A Steenbergen ras at e-gerbil.net
Sun Sep 27 05:00:14 EDT 2009

On Sun, Sep 27, 2009 at 04:44:48PM +0800, Shane Short wrote:
>  Hi Richard,
> The OP said he was getting the 9.2Gbps measurement from MRTG, which  
> would've been a 5min average, hence my comment. Your explaination  
> sounds much more plausible though :)

Oh yes I suppose he did, I just read the show interface line and
mentally blocked out the MRTG one. :) If that really is the case than
averaging is certainly going to be a factor. Honestly, MRTG is quite
possibly the worst graphing package anyone could possibly be using in
this day and age. Unless you just stepped out of a cryogenic freezer
from 1999, anyone still using MRTG should quickly find the nearest knife
and commit seppuku. Go ahead, I'll wait. :)

Also Juniper doesn't count layer 2 overhead in its reported counters
(unluke most other vendors), and sonet overhead is pretty healthy, so
this could easily account for a significant portion of the "missing
bits". I'd suggest he start with a monitor interface, and assume the 
rest is L2 overhead.

Richard A Steenbergen <ras at e-gerbil.net>       http://www.e-gerbil.net/ras
GPG Key ID: 0xF8B12CBC (7535 7F59 8204 ED1F CC1C 53AF 4C41 5ECA F8B1 2CBC)

More information about the juniper-nsp mailing list