For my experimental purpose i need to drop packets following poisson probability distribution and Markov Model. But i do not know how to do it. how to map the theory with the packet drop. if it can be done then the essenece of internet packet drop will be found. i will use it to formulate the performance of VoIP in user perspective.

A c kernel module,user level module, using iptables or any other way will work. If you know how to drop either by poisson or markovian that will also be ok. Can anybody help me?

You need to actually drop packets? In the kernel?!

Tell us more about your test-scenario, and what exactly it is that you are testing. It would be far easier to write your test program to randomly (or otherwise) decide to pretend that the packet it has just received was actually dropped, doing the whole thing in user-level code... say in a modified library or a "wrapper" library.

Certainly packet dropping. But that drop have to follow poisson probability distribution. Any level programming or even iptables will work but i need this dropping must follow poisson distribution. right now i donot need to know about markov model.

two persons will talk through microphones of pcs (soundcard is installed) and the two computers are connected through lan. Ip telephony software like openh323 is used. one of the pcs is running in linux.

A third linuxbox may be used for forwarding packets from where we can drop the packets. It will be easier. The forwarding linuxbox is not mandatory. 2 pcs may connect directly. We will drop packets from the linux client.

However the packet dropping fashion must follow the poisson distribution to get the essence of internet packet loss. Itis already established. Then we can measure how the voice quality will be in Internet Telephony

I'd start with Wikipedia...
http://en.wikipedia.org/wiki/Poisson_process

Any sort of user-level proxy process ought to do the trick.

Perhaps Dummynet ( http://info.iet.unipi.it/~luigi/ip_dummynet/ ) or NIST Net ( http://snad.ncsl.nist.gov/nistnet/ ) would be helpful to you.

Regards,
Jacob