Lecture 12

# In case of single pipe we have total poisson arrival

This preview shows page 1. Sign up to view the full content.

This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: ider m statistically identical and independent Poisson packet streams each with an arrival rate of =m packets per sec, transmitted over a communication link with exponentially distributed service time with mean service rate . In case of single “pipe", we have total Poisson arrival as m=m =  due to one of properties of Poisson process mentioned earlier. Using M/M/1 model, we get the avg. delay to be 1 Ta =  ,  : If, in, the link is divided into m separate pipes (partitioned), then the service rate of each pipe is =m, i.e., 1=m-th of the rate of the original pipe. In this case, the avg. delay (again using M/M/1 on one partition): 1 Tb = =m , =m =  m  : , Thus, splitting the line into m pipes increases the avg delay by m times! This is an interesting phenomenon about statistical multiplexing; as a rule of thumb, it is better to have a fatter link than partition the link into lower speeds since it results in more delay on average. What about the effect on average number of pac...
View Full Document

## This document was uploaded on 03/19/2014 for the course CS 6030 at Western Michigan.

Ask a homework question - tutors are online