1. Give examples of applications that generate elastic traffic.
FTP, email, web surfing.
2. Which QoS characteristics are the most important for time-sensitive traffic?
Delay, delay variation, loss
What are the
positive and negative effects of using queues in
Positive: A queue is a result of packet buffering, which aims smoothing of traffic burstiness. It means that unless a buffer is not overflown a network can accept packets even though rate of ingress traffic temporarily exceeds a network capacity.
Negative: However, buffering results in increasing of packet transferring time. On top of that a random character of packets generation results in random sizes of packet queues, which it its turn leads to random values of packet delays, which makes network behaviour uncertain and unpredictable. This feature of packet-switched networks has a negative impact on time-sensitive traffic, e.g. voice traffic.
1. Which parameter has the greatest influence on queue size? Which parameter is the second most important?
A utilization of a server is a primary factor; the second most important factor is a variation of requests arrivals at a server.
2. What kinds of traffic does the packet-switched network transmit? What are their requirements for the network?
3. What are the advantages and drawbacks of priority queuing?
The main advantage is that priority queuing guarantees a minimum delay and loss for highest priority traffic. The drawback is that lower priority traffic doesn’t have guarantees for dedicated bandwidth.
4. For what kind of traffic is weighted queuing most suitable?
It is most suitable for elastic traffic that needs bandwidth guarantees.
5. Is it possible to combine priority and weighted queuing?
6. Can packets of the highest priority be delayed in queues?
Yes if highest priority traffic utilises more than certain limit of a server capacity.
7. List the methods of congestion control and congestion avoidance.
What are the
differences of bandwidth reservation in packet-switched and
In circuit-switched networks, a bandwidth reservation means that the bandwidth is dedicated to some connection in full and for all duration of the connection and cannot be used by other connection even though the connection does not temporarily use a dedicated bandwidth in full. In packet-switched networks, it means that a reserved bandwidth can be redistributed dynamically if a connection does not temporarily use the bandwidth in full.
9. What are the components of the reservation-based QoS system?
(1) Queuing mechanisms;
(2) Recourse reservation protocol;
(3) Traffic conditioning mechanisms.
10. What problem is solved by traffic engineering methods?
Traffic engineering (TE) methods solve the problem of choosing routes for flows (or traffic classes) and accounting for QoS requirements.
11. What traffic parameter is changeable in traffic engineering?
1. Suppose that some data flow belongs to the CBR class. Data are transmitted in packets equal to 125 bytes through a 100 Mbps link. The traffic profile has the following parameters: PIR for burst periods is 25 Mbps, maximum deviation of the interpacket interval is 10 ms, and the burst period is 600 ms. If the traffic conforms to its profile, what is the maximum volume of the burst?
First, let us find an average inter-packet interval T on a burst period:
T = S /PIR, where S is a packet size. T = 1000 bits / 25 Mbps = 40 μs.
When deviation from the average inter-packet interval is maximal the instant value of inter-packet interval is 30 μs. The information rate PIR = 1000 /30 = 33.3 Mbps. Maximum burst size Bmax = 33.3 x 600 = 20 000 bits.
2. Which of the five flows will on average spend less time in the queue to the output interface 100 Mbps if these flows are served by weighted queues and are allocated 40%, 15%, 10%, 30%, and 5% of the interface bandwidth, respectively? Average flow rates are 35, 2, 8, 3, and 4 Mbps, respectively. Coefficient of variation of interpacket intervals is the same for all flows.
Flows will be served with utilisations
35/40 = 0.875; 2/15 = 0.133; 8 / 10 = 0.8; 3 / 30 = 0.1; 4 / 5 = 0.8 respectively:
Fourth flow will experience the least queue delays as it will be served with utilization 0.1 which is the least one among all flows.
3. For which of the events listed here might the flow from the queue with the highest priority have to wait in a queue?
A. Queues of lower priorities
B. Its own burst
C. Bursts of low-priority traffic
Its own burst
4. There are three queues to the output interface 2 (ATTENTION: it is a typo in the book’s text, which says 10) Mbps, served according to the weighted queuing algorithm. There are three packets in the first queue, packet 1 is 1,500 bytes, packet 2 is 625 bytes, and packet 3 is 750 bytes. In the second queue are packet 4 (500 bytes), packet 5 (1,500 bytes), and packet 6 (1,500 bytes). In the third queue, there are packet 7 (100 bytes), packet 8 (275 bytes), packet 9 (1,500 bytes), and packet 10 (1,500 bytes). In the queues, the packets are in ascending order (i.e., the first packet in the first queue is packet 1; in the second queue, it's packet 4; and in the third queue, it's packet 7). In what order will the packets appear at the output of the 2 Mbps interface, if the working cycle of the algorithm is 10 msec, and queues are allocated 50%, 30%, and 20% of the resource bandwidth, respectively? In each cycle, the algorithm always takes a packet from the queue (provided that it is not empty), even if the packet size guarantees that its transmission will exceed the time allocated for this queue.
The output packet sequence will be:
1, 4, 5, 7, 8, 9, 2, 3, 6, 10
5. How much time would it take to complete each of the two cycles of processing queues (see the previous question)? At what rate was each flow served at this interval comprising two cycles?
The first processing cycle consists of serving of packets 1, 4, 5, 7, 8 and 9, respectively; packets have total size of 12000+4000+12000+800+2200+12000 = 43000 bits. Hence, T1 = 43000 /2 x 106 = 21.5ms.
The second processing cycle consists of serving of packets 2, 3, 6 and 10, respectively; packets have total size of 5000+6000+12000+12000 = 35000 bits. Hence, T2 = 35000/ 2 x 106 = 17.5 ms.
6. How is it necessary to change the cycle time of the algorithm described in question 4 to make flow rates closer to the planned ones? Increase or decrease?
7. At the network ingress, some flow is undergoing policing according to the 3 Mbps profile. This flow is allocated 30% of the 10 Mbps output interface bandwidth in the transit network switch. Which of the statements here are correct?
A. The result of applying either of these mechanisms is the same; therefore, it is not necessary to implement reservation in the switch.
B. The result of applying either of these mechanisms is the same, but the reservation in the switch is necessary because at the network ingress and within the switch, the flow competes for resources with other flows.
C. The results of applying either of these mechanisms are different. At the network ingress, the flow rate is limited by 3 Mbps; in the switch for this flow, the rate of 3 Mbps is guaranteed even during periods of congestion. - Correct
8. Is it possible to have no queue in a system whose utilization coefficient is close to 1?
Yes, it is possible when ingress flow and serving time are deterministic, i.e. intervals between requests arrivals are equal and serving times are equal as well.
9. Which of the mechanisms listed here is necessary to use to ensure the high-quality transmission of voice traffic (64 Kbps stream) using the packet-switched network?
A. Reserve 64 Kbps bandwidth on all switches along the route of the voice traffic. Yes
B. Serve this flow in the priority queue on all switches along the traffic route. Yes
C. Use the input packet buffer on the receiving network node. Yes
D. Smooth the traffic in the output queues of all switches along the route. No
10. Is the following statement correct? Resource reservation in packet-switched network deprives the user of possibility of dynamically redistributing the bandwidth between flows. No, it is incorrect.
11. What mechanism must be applied to avoid the suppression of low priority traffic by traffic of high priority? Weighted queues.