[c-nsp] easy question, PPS v Megabits

[Mikael Abrahamsson]

On Tue, 22 Aug 2006, jim bartus wrote:

> Since others have covered the huge variance, here's my anecdotal evidence.
>
> Today while our WAN pipe hit 40mbps-in/200mbps-out it was doing
> 24kpps-in/32kpps-out.  This is for overwhelmingly HTTP traffic, with some
> real streams and ftp stuff too.  So short of being some sort of large file
> mirror service, this is roughly a best case scenario.

For real world calculations I'd say calculating with average packet size 
of approx 300 bytes makes a lot of sense, this gives some headroom as the 
actual internet mix is probably a bit higher in average size.

Agilent proposes ~340 bytes as average size in this document and comes 
with some claims regarding what sized packets are common on the internet. 
Their numbers seem to make sense.

<http://advanced.comms.agilent.com/n2x/docs/journal/JTC_003.html>

Otoh if you want to be really sure and make sure you'll be able to handle 
any kind of traffic (DDoS for instance) then you'd better go for the 
theoretical worst case which has been explored earlier in the thread.

-- 
Mikael Abrahamsson    email: swmike at swm.pp.se