INLS 184: Protocols and Network Management
March 21st – April 5th, 2005
Instructor: Jim Gogan

Mid-Term Examination

1) Define each of the following terms from a networking perspective:
    - latency
    - jitter
    - congestion
    - throughput

2) Arrange the following headers below in the correct order that you would find them in an Ethernet frame.
A.  FTP header            B.  TCP header          C.  Ethernet header          D.  IP header

3) In your readings, you were told that the IEEE 802 committee chose to "subdivide" the Link Layer (level 2) for networks into two sublayers: the MAC (media access control, or how a station actually gets to transmit data onto the physical layer) and the LLC (logical link control, or how multiple "higher level" (or more software oriented) protocols share the use of the data link. Describe the media access control PROCESS (i.e. how does a station transmit AND receive) on Ethernet.

4) Within an Ethernet environment, define the concept of "duplex".  What type of components in the Ethernet world, if any, are incapable of running at full duplex and why?  What would make you think you had a "duplex mismatch" between two Ethernet devices and would such a mismatch cause any problems?  Would two devices with an Ethernet "duplex mismatch" even be able to communicate with each other?

5) Briefly describe the differences (if any) among the 802.11 standards a, b and g; assuming I wanted to provide wireless coverage to an auditorium that seats 1000 people, what issues would there be if I just installed one 802.11b access point?  what issues would there be if I installed 20 access points in that same space? What if I used 802.11a for that situation instead?

6) Suppose a 10 Mb/sec Ethernet hub (repeater) is replaced by a 10 Mb/sec switch on a network where ALL traffic is between a single server and some number of client workstations.  Because all traffic must still traverse the server-switch path (to get to the server), nominally there is no improvement in bandwidth.   HOWEVER, would you expect ANY improvement in bandwidth?   If so, why? What other advantages and/or drawbacks might a switch offer versus a hub?

7) Let's say that we have four networked workstations (X, Y, Z, and W) and learning bridges B1, B2, and B3 with EMPTY FORWARDING (SAT) tables; i.e. the bridges JUST booted up and no one's transmitted any traffic yet (see picture below).

• if X sends a packet to Z: which bridges learn where X is, and does Y's network interface see this packet?
• if Z now sends to X (after X has sent to Z): which bridges learn where Z is, and does Y's network interface see this packet?
• suppose Y now sends to X (after the above two packets): which bridges learn where Y is, and does Z's network interface see this packet?
• finally, suppose Z sends to Y (after the above three packets): which bridges learn where Z is, and does W's network interface see this packet?

8) How does a computer (technically speaking, its Ethernet driver and operating system) know whether an arriving frame contains an IP datagram or an ARP message?

9) Say that you want to start a telnet session between your desktop computer (mine.my-dept.unc.edu) and another computer on campus (neat-stuff.unc.edu). Even though we haven't discussed TCP/IP applications yet, telnet is a remote login application that uses TCP as its transport layer protocol and IP as its network layer protocol. Describe thoroughly the network traffic (i.e. each packet from all hosts/computers) that results BEFORE THE FIRST TCP PACKET (read that phrase in caps carefully!; remember you can have IP packets without having a TCP layer) itself ever goes out on the wire. Assume that you've just booted your machine so you have no address cache for any computer on the network; also assume for this example that the two machines are on the same "logical" network; and finally, assume that your computer is already configured with a static IP address (i.e., you're not using DHCP).

10) In Assignment #2, you were given a series of "byte-level" bridge filter problems, complete with instructions on how to create these bridge filters. The problem below uses the same concept and instructions; it just gets at a higher layer in the protocol layers than the previous problems. Remember, that we are dealing with TCP/IP over Ethernet in this problem; therefore, this will be Ethernet II. You're looking for packets with %0800 in byte %C, and the IP header starts at byte %E. Remember how to count in hexadecimal!

You are seeing with a protocol analyzer that somebody has decided to use one of your machines as a "ping" target and is constantly putting continuous 1000 byte "ping" packets onto your network. You've decided to keep all ICMP/ping "echo request" packets off of your network. (Assume that the length of the IP header is 20 bytes; i.e., no "IP Options"/padding.) Remember, a "ping" is an ICMP echo request.

11) Describe EVERYTHING you can about EITHER ONE of the following two packets (where "xx xx xx xx" indicates a valid CRC/checksum).   You'll need to use the knowledge you've acquired about what's actually in an IP header to fully decode these packets.  I want more than just the information that was asked for in the first homework assignment (i.e. I expect a lot more information than just hardware vendor, frame type, and network protocol).  Aside from the specific details of what's in each byte location, what's that packet really telling you as a network administrator? i.e. INTERPRET THE PACKET (don’t just decode it)

02 60 8c 2c e8 dd 00 80 96 00 97 77 08 00 45 00
00 38 15 11 00 00 ff 01 1f 88 98 02 41 dc 98 02
15 4b 05 01 f0 90 98 02 15 ff 45 00 00 70 c7 72
00 00 1d 11 00 00 98 02 15 4b 98 02 15 ff 00 7f
00 7d 00 48 d5 e4 xx xx xx xx

aa 00 04 00 35 9d aa 00 04 00 d6 9d 08 00 45 00
00 38 00 00 00 00 fe 01 66 a8 98 02 11 01 98 02
15 17 03 01 72 b4 00 00 00 00 45 00 00 54 ec 31
00 00 1c 11 e7 c2 98 02 15 17 80 6d 9d 1e 08 47
00 6f 00 40 81 54 xx xx xx xx

12) One of the most important aspects of developing any reliable data transfer protocol that depends on appropriate acknowledgements to verify that packets have been received and are ordered properly (like TCP) is the use of timers, to be able to assume that if an acknowledgement hasn't arrived within that time, it must have gotten lost.   However, as Perlman notes, actually determining an appropriate value to use for the timer can be tricky.  What are the problems associated with having a timer value that's too long?  What are the problems associated with having a timer value that's too short?   Can you just set a timer value at the start of a session between a source and destination and use that original value for the entire length of the session?   If not (that should give you a clue as to what the answer of that last question is), why not?

13) We've talked about two different transport protocols available within the IP world: TCP (Transmission Control Protocol) and. UDP (User Datagram Protocol). What are the differences between these two protocols?  Why do we need two different transport protocols?   Are there applications for which TCP is better suited than UDP; if so, why?  Are there applications for which UDP is better suited than TCP; if so, why?

14) Consider a router that uses a mechanism that responds to congestion early by dropping OCCASIONAL packets before its queue is full, instead of the normal process of waiting until the queue is full and then drops all arriving packets. The first router is likely to cause some senders to lose a single packet from a congestion window’s worth, while the second router will cause senders to lose many packets from a single window – perhaps even the entire window’s worth. Does this mechanism exist on the Internet today; if so, what is it? Also, explain why the first router’s kind of loss is likely to be detected by the sender more quickly than the process used by the “normal” router.

15) In our discussions on TCP, we talked about both flow control and congestion control.

a) What is the difference between flow control and congestion control as used in TCP?

b) What does TCP use to provide FLOW control?

c) Describe at least two mechanisms that TCP uses to provide CONGESTION control?  (Don't just say what they are; DESCRIBE what they do, how they work, and how they help control congestion.)

d) Are there any ways that a sender can become aware of congestion WITHOUT packet loss?

16) Duplicate acknowledgements in TCP can be a sign of a lost packet:

a) How many duplicates are needed before TCP will assume a packet was lost?

b) What happens to the congestion window size when a lost packet is detected in this way?

c) If a lost packet is detected in this way, will the sender then go into slow start mode or do something else?

17) What is a VLAN?  Under what circumstances would using VLANs prove beneficial, and why?  What, if any, standards exist for VLANs?  Are there any issues associated with having a "VLAN-capable/configured" switch directly connected to a switch that was not "VLAN-aware"?  If so, what?

18) Assume two routers are (unintentionally - we hope) misconfigured to form a routing loop for a destination X.  Explain why an IP datagram destined for X will not go back and forth forever.

19) Let's say that Dept. X is currently connected to the campus network through a Layer 2 switch.  They have been assigned the "subnet" 65 in the network 152.2.0.0 that has been assigned to the entire UNC-CH campus.  Since the deptartment was bridged to the campus network, even though all of their addresses are in the 152.2.65.0 network, they have been (correctly) using the netmask of 255.255.0.0 and a default route address of 152.2.254.254 (the address of the main campus router connected to the outside world), since there are no other routers between that deptartment and the main campus router.

Now, let's say that the departmental network has gotten so big and busy that they want to segment it into FOUR separate broadcast-domain networks.  To do so, they have asked to have a Layer 3 switch (router) installed to isolate all that broadcast traffic and to create the separate broadcast domains.   This means that they need a five port switch; one port to connect to the backbone, and four ports to connect to each of those separate broadcast domains within the department.

KEEP IN MIND, HOWEVER, THAT EACH PORT OF A ROUTER (UNLIKE A BRIDGE) NEEDS ITS OWN INDIVIDUAL IP ADDRESS REPRESENTING THE FACT THAT EACH SEGMENT CONNECTED TO A ROUTER IS A SEPARATE IP NETWORK.  WHICH LEADS US TO .....

So, the department now has a five-port router (Layer 3 switch): Port A is connected to the campus network, and as such, has been assigned the IP address of 152.2.254.65 by the campus' Networking group.  Ports B-E need to be assigned addresses by the department, since they are inside the departmental network.  The department, as stated above, has the IP network 152.2.65.0 already assigned to them; they don't want to ask for (and wouldn't get, even if they asked) separate "Class C" IP networks for each segment to be connected to the router, since each segment will have 60 or fewer workstations connected to them.

Now .... finally .... here is the question:  How should the department configure the router and the IP addresses (and appropriate netmasks and default router address) for each of the four segments such that each of the four departmental segments has an equal number of possible hosts associated with it, and insuring that they are all part of the 152.2.65.0 range?  Under your approach, which of the addresses from 152.2.65.1 - 152.2.65.255 are not available for use by a workstation or a router interface (treated the same as a workstation) and why?

20)   Finally, bet you thought you were going to get away without a questions about the OSI Model, didn’t you? What is the "OSI 7-Layer Model"? PLEASE BE EXPLICIT in describing EACH of the 7 layers and what they do (or are supposed to do).  BRIEFLY describe what is meant by the concept of layering in networks. What are some advantages and disadvantages of layering?   And, finally, which of the 7 layers of the ISO OSI standard defines:
   Bridges?
   Format of Ethernet frame?
   Shape of the network plug?
   DNS?
   Internet Protocol?
Repeaters?
   Maximum length of a cable segment?
   End-to-end flow control?
   Routers?
   Transmission Control Protocol (TCP)?