[j-nsp] Accuracy of interface stats

[Richard A Steenbergen]

On Sun, Sep 27, 2009 at 02:52:34PM +0800, Shane Short wrote:
> You'll find that 9.2Gbps was the 5minute average-- you were probably  
> bursting well above that frequently, hence the drops.
> 
> There was a massive discussion about this on the nanog list, depending  
> on the type traffic your transiting, you shouldn't run your link at  
> hotter than 70-90%

I don't think the show interface counters are 5 minute averages. A
"monitor interface" updates every 2 seconds and seems to catch major
changes in traffic in about that much time, so the hardware is clearly
capable of updating counters within a few seconds (definitely well under 
10s). I don't spend much time looking at show interface counters, but a 
quick poke shows them updating what appears to be every 2-3 secs, so it 
seems likely that these would be similarly real-time to the monitor 
interface rate.

To the OP: This doesn't happen to be a OC-192 doing WAN PHY 10GE does
it? Those will definitely top out at almost exactly 9.2Gbps, after you
figure in the overhead for sonet + ethernet + the lower data rate.

-- 
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)