Department of Electrical Engineering

Communication Networks (EED304) - Monsoon 2022

Instructor: Prof. Vinod Sharma
ASSIGNMENT-4

1. Consider the following queue, where the messages are iid exponentially distributed with mean L
bits. Transmitter has a rate of c bits per sec (bps). The arrival rate is λ packets/sec.

(a) Give mean delay formula for λ = 10 bit/sec, L = 8 × 106 . Find the minimum c such that the
mean delay is ≤ 0.1 sec.
(b) How does the average delay change if c and L are multiplied by the same constant?
(c) Files are divided into K subfiles. The rate of request of a subfile is Kλ . Then, determine the
mean delay of the subfiles.

2. All packets are 1000 bits. Suppose A is a video traffic arriving at a link at a constant rate of 1
packet/msec (rate is 1Mbps). Suppose B is a highly bursty traffic with 10Mbps worth of packets
for every 10 sec (rate is 1Mbps). The scheduler at the link is FCFS (First Come First Serve). A
gets priority when both A , B packets arrive at the same time.

(a) Calculate the maximum packet delay of A packet.

(b) Calculate the maximum delay of A, when the scheduler is a Round robin scheduler.

3. Consider the following queue. The service times at the two servers are iid exponentially distributed
with parameters µ1 and µ2 .

(a) Give expression of total mean sojourn time D(x).

(b) Find x that maximizes f (x) = 10ln(x) − 2x such that D(x) ≤ 0.4 sec.

1
4. (a) Consider the following queue

Find its mean sojourn time under stationarity.

(b) Find the mean sojourn time for the following system.

1
Comparing with part (a), we see that in part (a), the mean sojourn time is 2 of this. What
may be the intuitive justification for this?

5. Consider the system given below as a transmission link which takes in transmission and propaga-
tion (of packet and ack) a total of time s. A transmission fails with probability p. Compute the
mean sojourn time for this queue (this queue can be considered as an M/GI/1 queue, compute its
service time).

6. In the following network, assume external arrivals are independent Poisson processes and service
times at each queue are iid exponential. Obtain the stationary distribution of the system.

7. (a) Consider the following queue with the Poisson process with a rate of 100 packets/sec, µ = 50
packets/sec. Find Lim P(qt ≤ x)? Argue that the output process of this queue will be Poisson
t→∞
with a rate of 50 packets/sec.

(b) Now, consider the queueing network shown below with µ1 = 125 packets/sec, µ2 = 150
packets/sec, and µ3 = 15 packets/sec. Compute Lim P(qt (i) ≤ x), i = 1, 2, 3.
t→∞

2
 8. (a) Consider a large buffer B = 50Mb. 10 users are feeding it, each with effective BW 2Mbps
and 5 users with effective BW 1Mbps each, for δ = 0.1. Compute the transmission rate
needed to transmit the data from this buffer such that the probability of overflow from the
buffer is ≤ exp(−0.1 × 50 × 106 ).
(b) In the setup of (a), let the decoupling BW of the first class of users (group of 10) is 1.5 times
that of the effective BW. Then what is the transmission rate of the link needed, so that flow 1
in the first group retains its effective BW at the output of the queue?
9. Consider a single source, single server queue. Let a(t) be the number of bits generated by the
source at time t. Let {a(t)} be iid Bernoulli such that
(
M with probability M1
a(t) =
0 with probability 1 − M1
) d Λ(δ )
Calculate effective BW Λ(δδ and show that dM δ > 0. This shows that effective BW increases
with the burstiness M of the source.
10. Suppose we have n1 traffic sources of type 1 using leaky buckets with parameters (a1 , B1 ), and
n2 sources of type 2 using parameters (a2 , B2 ). These all pass through a buffer of size x and
transmission link y. What should be (x, y) so that there is no buffer overflow?
11. Consider the following figure

(a) P(ON) = 0.2, rate of source during ON time is 1000 bytes per slot. Give a value (a, B) of the
Leaky bucket such that for B1 = 10Kb, the probability of overflowing B1 ≤ 0.001. Also let
every packet ≤ 100 byte.
(b) Find C when B2 = 5KB such that there is no packet drop in B2 .
(c) To support 10 such Leaky bucket controlled sources at Q2 how much C, B2 are needed?

12. Consider two TCP connections f1 and f2 as shown below.

3
 (a) Suppose only f1 is active and in collision avoidance phase (AIMD) with multiplicative factor
1
2 . The RTT is 50 msec. Calculate long-term average throughput of f 1 .
(b) Now, assume both f1 and f2 are active, both in collision avoidance phase. RTT of f1 is 50
msec, and f2 is 100 msec. Find long-term average throughput of f1 and f2 .
(c) Use log utility function. We want to maximize the total utility of the network. Compare this
optimal solution with the one obtained in (b).

13. TCP connections T1 and T2 in the following figure use AIMD: increases by 1 packet and decreases
1
by 1.1 . The rate is window size/RTT. The RTT for T1 is 0.1 sec and for T2 is 0.2 sec.

(a) Draw the diagram of evolution of (x1 , x2 ) with time, where xi is the rate of connection i, i =
1, 2.
(b) Find out the approximate limiting value of (x1 , x2 ).

14. Let (x1 , x2 ) be the rates of flows 1 and 2, respectively as shown in the figure below. The link (A,
B) has a capacity a and the link (B, C) has a capacity b.

(a) Find x = (x1 , x2 ) that maximizes x1 + x2 subject to capacity constraints.

(b) Find x that maximizes min{x1 , x2 } and satisfies capacity constraints.
(c) Find x that maximizes log(x1 ) + log(x2 ) subjected to capacity constraints.

15. Consider a cellular wireless network with one BS and two mobiles. The network can be in one of
two equally likely states c1 = (1, 2) and c2 = (3, 1). Channels states are iid across time, and BS
can transmit to only one mobile in one-time slot. Draw the capacity region of the network.