Hi,

I think that it's a normal behavior. The size of a non-drop queue 
depends on the minimum observed inter-arrival rate and the serving 
capabilities of the link layer. In your case, it seems that 200 packets 
offers enough room/space for your incoming packets, given a saturated 
server, to avoid any packet loss. If you increase you inter-arrival rate 
by say increasing the rate_ parameter for a CBR traffic you'll notice 
that you'll have to further increase the size of your drop-tail queue in 
order to avoid a drop. Increasing a queue size actually protect the 
flow's packets but also introduces more delay which is bad for 
interactive/real-time protocols. Anyway, real-world hardware (routers, 
LAN cards, etc.) doesn't have infinite buffer space.

Hope it helps,
Ilyes Gouta.

Ethan Giordano wrote:
> This is how I am setting the size of a droptail queue:
> 
>       $ns queue-limit $n1 $n2 $X
> 
> I am seeing behavior where for 1 <= $X <= ~180 everything works fine.
> 
> If you push X to say 200, then suddenly there are no more packet drops, 
> and there always appears to be 183-186 packets in the queue.
> 
> What I am trying to simulate is the rule of thumb buffer sizing which is 
> 2*BandwidthDelayProduct which for high speed links is on the order of 
> thousands of packets.
> 
> I've been through queue.cc, queue.h, drop-tail.h and drop-tail.cc and 
> can't figure out what's going on.
> 
> Any thoughts?
> 
> ~Ethan
> 
> 

Reply via email to