Chapter 19

Review Questions

1. When do software developers prefer to use UDP? When do they prefer to rely on TCP?

2. What volume of data (with a precision up to 1 byte) was received during the TCP session by the sender of the TCP segment whose header contained the value 1,845,685 in the ACK field, given that the first received byte had the number 50,046?

1,795,638 bytes

3. Is it possible to forward IP packets if the router lacks a routing table?

4. Is it possible to do without routing protocols in the network?

Yes, by building roting tbles manually.

5. What are drawbacks of distance vector routing protocols?

6. What is the main operating principle of routing protocols based on LSAs?

7. What is the difference between IGPs and EGPs?

IGPs are responsible for the choosing a route within an automomous system while EGPs choose a route between autonomous systems.

8. Which metric is used in RIP?

The number of hops between routers

9. Why does RIP consider the distance of 16 hops unreachable?

10. What methods are there for speeding up the convergence of RIP?

Split horizon, hold down, triggered updates - all these methods reduce a duration of periodds of instabiility and hence improve the RIP convergence

11. What are the main stages of building the routing table using OSPF?

12. What is the role of HELLO messages in the OSPF protocol?

13. What types of metrics does OSPF support?

OSPF metric can reflect the speed of network links; links reliability and delays.

14. For what purposes is the network of routers supporting OSPF divided into areas?

Computation complexity of OSPF sharply increases when dimension of a network increases. To cope with dimensioning OSPF uses areas. The routrers which belong to an area operate with a link database only of the area which decreases dimension of a link's graph.

15. What are the main disadvantages of OSPF?

Computation complexity

16. Why is EGP no longer used in the Internet?

Because there are loops beween autonomous systems in the modern Internet and EGP can't work prevent looping routes in this case.

17. What is the mechanism that allows BGP to operate in networks that have loops between autonomous systems?

18. What parameters are changed by the BGP router in the advertisement received from some autonomous system when it passes it to another autonomous system?

19. If a problem with an IP packet arises, in which cases is it impossible to send an ICMP error message?

ICPM messages are not generated when a problem with IP packet transferring another ICPM message arises. Such design decision was made to prevent storms of ICMP messages.

20. What is the destination of an ICMP message? Which software module processes it?

ICMP message is always sent to a node which was a source of a packet caused an error. ICMP message is processed by the operating system kernel or by protocols of the transport or application layers but not IP or ICMP themselves.

21. How does an ICMP message improve the reliability of data transmission in an
IP network?

ICMP improves the reliability of data transmission not by direct means like packet retransmission (like TCP does) but indirectly as it informs a source node about problems that arose during packet transmission accros a network.


1. Find a partner and model a TCP session. Negotiate the maximum segment size, the initial sizes of buffers, the initial values of the SEQUENCE NUMBER, and the window size. Then, start sending "segments" to each other asynchronously. The role of "segments" can be played by cards containing the values of the key fields — number of the first byte, the size of the segment being sent, the ACKNOWLEDGMENT NUMBER, and a new value of window size. You can pretend sometimes that you have lost the cards and act according to the TCP operating logic. Do not forget to mark the time on each "sent" copy of the segment to trace the arrivals of acknowledgments. This game will help you to better understand TCP. Feel free to ask questions.

2. What time, in the worst case, will elapse before the routing tables of the network shown in Fig. 19.15 come to a coordinated state after the R1 router loses connection to the network? Assume that all routers support the split horizon mechanism.

The incorrect (out-dated) information will be delivered to the router R1 by the router R4 which learned the information about the network from the routers R2 or R3 which depends on which of the two routers sent the information earlier; the metrics of the route are equal for both routers. Let assume that the router R4 has learned the information about the network in question from the router R3. In this case the router R4 applies the 'split horizon' rule for the network in question only to the router R3 and passes the incorrect information to the router R2 and this incorrect information ousts the correct one. It results in creation of a route loop which in worst case will exist for 4 x 30 = 120 seconds.

3. Suggest several variants of the metric that simultaneously takes into account the bandwidth, reliability, and latency of the communications link.