CS 640 - Final Exam
Fall 1996

ROUTING (2)
1. A routing protocol is said to be fair if it never directs a packet to an incorrect 
destination.
2. A centralized routing protocol depends on specially designated nodes to compute and 
then distribute routing tables.

DISTANCE VECTOR ROUTING (5)
3. Routing updates are sent by a switch only to directly connected switches.
4. Flooding is used to send each switchÕs routing table to every network switch. 
5. Hold-down makes it possible for a network to more quickly react to information that 
a link or switch has gone down.
6. Use of split horizon means that a switch does not send information about a 
destination to the switch from which it received its current route for the destination.
7. Distance vector is preferred to link state in networks with long paths. 

LINK STATE ROUTING (5)
8. The SPF algorithm no longer is used by link state protocols.
9. Each switch receives information about both ends of each network link.
10. The link state protocol requires O(E * N)  (O means order of) steps where E is the 
number of links and N is the number of network switches.
11. Link state is preferred to distance vector when there is minimal network 
congestion.
12. Hold-down plays a roll for link state analogous to its role in distance vector.

OSPF (8)
13. Different routes may be computed for different service types.
14. There may be multiple equal cost routes from a source to a destination.
15. When areas are used, each packet goes through an area border router.
16. When areas are used,  SPF is used separately on each area except the backbone.
17. Acknowledgments are sent on a hop by hop basis for link state updates.
18. Routes are re-computed when network congestion is detected. 
19. TCP is used to send update messages.
20. Information on external routes is flooded to all switches in the AS 
regardless of area.

BGP4 (8)
21. A BGP Speaker cannot use source routing to control the route of its packets 
to a destination.
22. A BGP Speaker can only advertise to its peers those routes that it itself uses.
23. Routes are usually optimal with respect to number of AS s traversed.
24. A reliable transport protocol is used to send UPDATE message.
25. The value of an AS_PATH attribute includes a sequence or set of AS s.
26. An  AS can have only one BGP Speaker.
The following may be included in a UPDATE message: 
27. One or more routes that have been withdrawn by the sender.
28. One or more routes for destination AS s.

ATM (13)
29. Switched virtual circuits differ from permanent virtual circuits in that the 
former supports throughput and delay bound guarantees.
30. AAL5 includes a length field in the packet header.
31. The AAL5 SAR layer consists of a single field in the ATM header.
32. When IP is used above ATM, some ATM cells may have total length more 
than 53 bytes.
33. In AAL5, data is followed by a trailer but there is no header.
34. With AAL3, the ATM checksum protects the full cell.
35. AAL5 has mechanisms to provide guaranteed bit rates.
36. AAL1 has mechanisms to provide guaranteed bit rates.
37. The ATM header has one bit to indicate cell priority.
38. If Token Bucket is used, every cell entering the network must be matched with an 
available token.
39. Token Bucket limits the size of bursts to the number of available tokens when the 
first cell of the burst is transmitted.
40. AAL5 adds a CRC and other information in a header.
41. The ATM header checksum ensures that cell corruption can be detected.

BRIDGES (5)
42. When bridges are used, optimality of path length is not a goal.
43. Bridges are often preferred to routers because they scale better than the 
latter.
44. Bridges are used to route IP datagrams.
45. A disadvantage of bridges is that always broadcast an incoming packet on 
every line except the one on which the packet arrived.
46. The root of the spanning tree used for routing is the bridge with the lowest 
ID.

IPv4 (15)
47. The IPv4 header is of fixed length.
48. An IP datagram is discarded if it reaches a network which does not support packet 
payloads of its length. 
49. If an IP datagram fragment does not arrive at the destination, then the fragment, 
not the entire datagram, is retransmitted by the source.
50. Subnetting is commonly used with Class B networks.
51. CIDR is used to make routing more efficient by aggregating routes.
52. The subnet mask is usually a subset of the network field of a class B IP address.
53. Subnetting is used to aggregate routes.
54. Each local area network can have multiple subnets.
55. A subnet can span multiple local area networks.
Class C networks can have 512 hosts.
56. IP datagrams have a checksum field to insure that the datagram has not been 
garbled in transit.
57. The IP header changes each time it passes through an IP router.
58. IP assumes that TCP is run on top of it.
59. An IPv4 option provides support for multicasting.
60. When an ARP query message is broadcast on the network, every host on 
the LAN adds the sender's link-level and IP address to its ARP table.
61. An ARP response provides a mapping between the IP anddress and the 
link-level address of the sender.

ICMP (5)
62. ICMP runs over IP.
63. ICMP Datagram Damaged is sent whenever an IP checksum error is 
detected.
64. ICMP Time Exceeded can help to detect routing loops.
65. The Internet Ping feature is supported by the ICMP Echo and Echo Reply types.
66. Traceroute uses the ICMP  redirect and time exceeded types.

IPv6 (6)
67. The IPv6 header is of fixed length.
68. To support compatibility with IPv4, an IPv6 address can have either 32 or 
128 bits.
69. One of the features of IPv6 is support for reliability.
70. One of the features of IPv6 is support for security.
71. One of the features of IPv6 is support for ATM switching.
72. IPv6  can have multiple headers.

MULTICASTING (6)
73. With Link State multicasting, each router maintains a single routing tree for each 
active multicast group.
74. With Link State multicasting, each router maintains a routing tree for each sender 
and each active multicast group.
75. With RPB, on each LAN there is one router designated to broadcast packets for all 
multicast groups and a given source. 
76. RPM is not as efficient as RPB in keeping networks that have no host in a group G 
from receiving packets destined for G.
77. With any of the studied multicast methods, the number of copies made to send a 
packet to group G never exceeds log n (base 2) where n is the number of members of G.
78. RPM prunes the set of networks that receive packets for G by excluding that have 
no hosts that are members of G.

DNS (6)
79. Recursive name server lookups are more common than iterative lookups.
80. There are one or more name servers for each zone.
81. MX records provide the domain name of a host that accepts email for a 
specified domain.
82. DNS maps domain names into IP addresses and IP addresses into Ethernet or 
other local area network addresses.
83. A client trying to resolve a domain name first accesses a root name server.
84. DNS lookup results are cached at hosts and can lead to misdirected IP 
datagrams.

VIRTUAL CLOCK  /  QUEUEING (5)
85. Virtual Clock warns senders if their virtual clock lags significantly behind 
real time.
86. Virtual Clock discards packets whenever a flow exceeds its negotiated 
average rate. 
87. Virtual Clock provides delay bound guarantees for flows that request this 
feature.
88. Virtual Clock implement a queuing discipline in switches.
89. A problem with virtual clock is that it allows a sender to gain an advantage 
over other senders by not sending for an extended period.

TCP CONGESTION CONTROL -  Congestion Window / Slow Start (6)
90. After reaching the receiverÕs advertised window, the send window is 
doubled for each round trip time.
91. The congestion window is set to half the send window when a timeout while 
waiting for an ack occurs.
92. Fast Retransmit results in retransmission of a packet upon receipt of 3 
duplicate acks.
93. After Fast Retransmit is invoked, Fast Recovery cuts the slow start period in 
half.
94. Slow Start features an increase of 1 in a TCPÕs send window for each round 
trip time.  
95. During Slow Start, a TCPÕs send window continues to increase until the 
receiverÕs window is reached or there is a timeout while waiting for an ack.

























4