Calculation Expected Number of Customers in an M/E_2/1 Queue

358 Views Asked by At

Following situation: I look at a queueing system with Poisson arrivals with rate $\lambda=4/15$, each arriving customer has to go through r=2 phases. In each phase, the service time is exponential distributed with rate $\mu =1$. Customers are served in order of arrival.

I should calculate the mean number of jobs in the system. I already computed the distribution of the number of uncompleted phases $$ p_n=\frac{7}{24} \Big( \frac{2}{3}\Big)^n+\frac{7}{40} \Big(-\frac{2}{5} \Big)^n $$ as well as the distribution of the number of customers in the system: $$ q_n=\frac{35}{48} \Big( \frac{4}{9}\Big)^n-\frac{21}{80} \Big(-\frac{4}{25} \Big)^n $$ where I used for the latter that $$q_i=\sum_{n=(i-1)r+1}^{ir} p_n$$.

My problem is to compute the mean number of customers in the system. In the solution to this exercise, it just says $$\mathbb{E}[L]=\sum_{n=0}^{\infty} n q_n=104/105$$. Putting my solution for $q_n$ into Wolfram Alpha gives exactly this result, but how do I actually compute this "by hand"?

(This question is about exercise 30 resp. 33 of the book "Queueing Systems" from Ivo Adan and Jacques Resing.)

1

There are 1 best solutions below

2
On BEST ANSWER

The analytical approach is to use the fact that $$\sum_{n=1}^\infty nr^{n-1} = \frac{1}{(1-r)^2}.$$ You'll have to write $q_n$ in the form of $r^{n-1}$ to do this.

The numerical approach is just to calculate the first $N$ terms of the sum, for some $N$. After some point, the terms $nq_n$ are going to get very, very small, and you can start to ignore them.