Re: [ns] goodput calculation?

[Brian Bowers <blbowers@cs.wisc.edu>]

Xiaojun Cao wrote:
> 
> Hi, ns-users,
> I have read all previous discussion on Good Calculation? But I am still
> confused which is the correct method.  I just want to calculate the goodput
> for the whole system. So it should exclude the duplication packet which
> maybe is a retransmission. Can I get this in following way:
>     Parse trace file, goodput=sum (all received TCP packets) - sum (received
> TCP packets have flag of "---A---")
>     received TCP packets:  TCP packets whose event is "r "

Simple, imprecise definition of GOODPUT
	number of unique TCP packets received

The number of unique packets received is simply the highest sequence
number received.  You can parse through a trace file to find the largest
unique sequence number.

In pseudo-code, it looks something like this:
  if this is a receive event then
    if the hop destination is the packet destination then
      if the current sequence number is larger than the held sequence
number then
        update the held sequence number

I've had trouble getting Perl to open large (3 GB) trace files, so you
may need to write this as a "filter".  Let the script process the
tracefile output as it is being generated rather than after the fact.

-- 
Brian Lee Bowers	|	CS 302 Lecture Sections 3, 4, and 7
blbowers@cs.wisc.edu	|	Instructor
http://www.cs.wisc.edu/~blbowers/cs302/index.html