COMS 6181 (Advanced Internet Services) Midterm Exam

Spring 2015

This is a closed-book exam, that is, no notes, books or other materials are permitted. However, you are allowed and encouraged to use a calculator. The exam has a total of 100 points, indicated in brackets. You have 110 minutes to complete the exam. Your blue book MUST contain all answers in numerical sequence, no more than one problem for each page, to avoid missing problems during grading.

  1. 7What fundamental common engineering design challenge do electric grids, packet networks, phone networks and (non-rural) roads all face?
    All of these engineered systems have a very high peak-to-average load, whether that's a hot summer day, busy hours on packet networks, Mother's Day or a natural disaster on (old) phone networks, and rush hour on non-rural roads. None of these systems can be economically engineered for peak load, but should avoid congestion collapse during those times.
  2. 9Suppose we replace today's broadcast AM, FM and over-the-air TV technology with IP-based delivery, using the same radio spectrum. What would be the major advantages and drawbacks or challenges, based on what we discussed in class?
    The advantage would be (possibly) more efficient delivery of content since newer video or audio encodings could be used. Also, it would be easy to multiplex many streams, in different audio and video quality, on the same channel. The same channel could be used to deliver data, or still images, not just audio and video. (As an aside, this was proposed very recently by somebody from the National Association of Broadcasters and there is a proposal for a satellite-based version for developing countries.) The major drawback would be the difficulty of accommodating all the legacy receivers on the same spectrum at the same time during the transition. Also, radios, in particular, would likely be more complicated than AM and FM radios.
  3. 6What's the difference between continuous media and streaming media? Given an example for each.
    Continous media describes all Internet-delivered media streams whose rate (bandwidth) is, on average, the same as the encoding rate. Streaming media is a sub-category where longer delivery or buffering delays can be tolerated.
  4. 6We have seen three fundamental interaction patterns. For each, name one application and one protocol (or indicate that the protocol(s) are proprietary to the application.)
    • Request-response: HTTP (e.g., web browsers)
    • Continuous media: streaming video using HTTP or proprietary protocols such as Flash video
    • Events: instant messaging using XMPP, Skype, Facebook or maybe email
  5. 6You are asked to design a new Internet (layer 3) protocol that includes a number of optional information elements. What kind of serialization would you use and why?
    Fixed length-elements and TLV for extensions, since that choice is easiest to parse, including in hardware, and compact.
  6. 9Show how you would encode a simple calendar entry (date, location, subject) in TLV, XML and JSON. (Rough syntax is sufficient.)

    For TLV, some hex or graphical representation, as in RFCs, is sufficient.

    [01] [10] "2015-03-23" [02] [02] [08] "New York" [03] [04] "exam"


    <location>New York</location>


      "calendar": {
        "date": "2015-03-23",
        "location": "New York",
        "subject": "exam"
  7. 6In the Internet protocol stack, what layer(s) ensure reliability?
    Transport layer such as TCP and, in some networks, the link layer (e.g., for Wi-Fi). In some cases, the application layer may also provide reliability if an unreliable transport protocol such as UDP is used.
  8. 6Both virtual circuit and packet switching use small data units (called packets, cells or frames, depending on the technology). What is the difference between the two?
    Packet switching does not set up any per-flow or per-session state in routers/switches, [such as label mappings for MPLS]. Virtual circuits do. Virtual circuits switch packets based on flow labels, while packets are routed based on global IP addresses.
  9. 6A crosspoint switching network can be either buffered (switching fixed-sized cells or frames, with memory at each cross point) or unbuffered (as depicted in class). Why would you add memory to a crosspoint switching network? What are the major advantages and disadvantages of each approach?
    Adding memory reduces blocking since intermediate nodes can buffer data when the output link is busy. Thus, the number of switch points can decrease.
  10. 6A speaker produces an audio signal of 10 Watts, and the background noise is 1 mW. What is the SNR in dB?
    The ratio is 10 / 0.001 = 10,000, which corresponds to 40 dB.
  11. 7A researcher has invented a new transport protocol. Outline what all needs to change for the protocol to be widely deployed and some of the obstacles. What are the typical steps this type of technology traverses before it becomes a "product"?
    • The solution has to be standardized, through some kind of standardization body.
    • One or more operating systems (e.g., Linux or Windows) will have to implement it, likely in the OS kernel.
    • Extensive testing under a variety of operating conditions.
    • Applications will have to implement or use the protocol.
    • Middleboxes such as firewalls will have to allow the protocol.
  12. 6What are the three largest components of mobile Internet traffic? (There's no need to estimate percentages or absolute values.)
    According to the Cisco data, video, web and mobile audio. [See slide 22 on 3/2/2015.]
  13. 8A fiber can carry 10 Mb/s of packet data. How much data can be carried across that fiber in a month? Is your theoretical answer realistic? Justify your answer.
    107 * 3600 seconds/hour * 24 hours/day * 30 days / 8 = 3.24 1012 or 3.24 TB. The number is unrealistic since no packet network can be loaded to 100% of capacity. Alternatively, it also ignores packet header overhead and any packet loss.
  14. 6"All other things equal, increasing the power (measured in Watts) of a cellular radio signal increases the achievable transmission rate." True or false? Justify your answer.
    True. Increasing the power increases the signal-to-noise ratio and thus, by Shannon, the achievable transmission rate. [The increased SNR allows a higher-rate modulation or reduces the bit error rate, and thus the effective transmission rate. The question is somewhat ambiguous since it uses the imprecise term transmission rate rather than information rate or net data transfer rate.]
  15. 6What is a delay-tolerant network? Name two applications where such a network may be useful.
    A delay-tolerant network continues to function even when the transmission path has extremely long delay or no contiguous connectivity. Examples include inter-planetary communications (e.g., from Earth to the outer reaches of the solar system), locations without 3G/4G cellular services or hard (or expensive)-to-reach places in developing countries. Services that may use delay-tolerant networks include message delivery (email), backups or content delivery (e.g., movies or educational content).

Last updated by Henning Schulzrinne