Computer Networking • CS 514

Test 4, 26 March 2014

This test has six questions. The test runs the entire class period, starting at 4:30 and ending at 6:20. When you finish the test, you may leave.

  1. Keeping in mind that giving an example of X is not defining X, briefly (in one or two sentences, 25 to 30 words) define:

    • Channel partitioning protocols

    • Generator polynomial

    • Switch

    • Byte stuffing

    • Point-to-point link

    • Channel partitioning protocols:
    A protocol that shares a link among several hosts by assigning some fraction of the link’s capacity to each host.

    • Generator polynomial:
    The divisor polynomial used in CRC.

    • Switch:
    A layer-2 device that connects several LANs (and the occasional directly connected host) into a single larger LAN.

    • Byte stuffing:
    A technique that neutralizes protocol-significant flag-byte values in payloads by preceding the values with an escape-byte value (which also must be neutralized when it occurs in payloads).

    • Point-to-point link:
    A connection between two hosts.

  2. Explain a problem VLANs solve, and the way in which VLANs solve the problem.

    Switches and LAN segments combine to form a single large LAN. Occasionally it’s useful to maintain the structure of the LAN segments as part of the larger LAN. For example, maintaining LAN-segment structure can support per-LAN management and security policies, and isolate traffic to specific LANs. VLANS can be used to maintain LAN-segment structure.

    VLANs work either by assigning specific switch ports to particular LAN segments or by marking each frame with a VLAN identity. In either case switches keep track of VLANs defined and makes sure traffic on a particular VLAN is kept to the resources devoted to that VLAN.

  3. Explain how the point-to-point protocol (PPP) might be used in a switched-Ethernet network. Explain why PPP would not be used in a switched-Ethernet network.

    Switched Ethernet has a star architecture: the switch in the center and the hosts (and other switches) around the periphery. Each link in the star architecture is a point-to-point link, and PPP can be used to provide communication over the link.

    PPP would not be used to provide communication over the links in a star architecture because Ethernet is also a point-to-point link, and PPP isn’t needed.

  4. Does an ARP table need to keep full IP addresses, or can it just store the network portion of the IP address? Justify your answer.

    An ARP table needs to keep the full IP address. ARP maps IP addresses to MAC addresses. If the ARP table kept only the network part of an IP address, the table couldn’t unambiguously associate a MAC address with a particular IP address.

  5. A trit (trinary digit) can assume one of three values (high, medium, or low, for example, or -1, 0, or 1). Suppose a network represents data using trits (the particular scheme the network uses doesn’t matter). Describe a simple parity-based error detection scheme for frames in this network. Give an example of an error not detected by your scheme.

    Perhaps the easiest parity scheme for trits is to keep track of parity for one of the trit values. For example, even-0 parity would use an extra trit to make sure any trit string has an even number of 0s. This scheme fails to detect any errors in the other two trit values, and doesn’t detect an even number of errors in the protected value.

    A slightly more useful parity scheme keeps track of the parity of all three trit values. For example, even parity uses an extra three trits to make sure any trit string has an even number of trits for each value. This scheme detects an odd number of errors for each trit, but doesn’t detect an even number of errors.

    There are lots of other possible parity schemes.
  6. Suppose you wanted to create a smaller Ethernet frame by removing a single field. What field would you remove? The resulting smaller frame should not alter the way in which Ethernet operates.

    Ethernet provides an unacknowledged datagram service, so there’s no need to keep track of the sender’s MAC address because it’s never used. This answer assumes the data-link layer is part of a larger protocol stack. If the data-link layer is the top of the protocol stack, then the sender’s address most likely has to remain in the frame (unless the application PDU includes the sender’s address, or implements a protocol where the sender doesn’t matter).

This page last modified on 2014 March 31.