Q1Circuit Switching vs. Packet Switching Slides 30–36

Names:

Four bursty users — A, B, C, D — share a link that supports 2 circuits. Each is active only 20% of the time. Under circuit switching, A and B each get assigned a circuit for the full 30 seconds.

The grid shows when each user has data (an X), one column per second.

12345678910 11121314151617181920 21222324252627282930
A
B
C
D
Black cell = that user has data during that second.
  1. A holds a circuit for all 30 seconds. In how many of those seconds does A actually send, and what percentage of the circuit sits idle?
    A sends in: seconds Idle: %

  2. In second 5, D has data and A is silent. Can D use A's idle circuit? Over these 30 seconds, what happens to C and D's traffic?

If the same link is packet-switched instead, nobody reserves anything and all four users transmit whenever they have data.

  1. In which seconds does demand exceed the link's capacity of 2? What fraction of the 30 seconds is that, and what must the link do with the traffic it cannot send immediately? What happens if too much traffic backs up?

    Seconds: Fraction: %
  2. Count the seconds' worth of user data the link actually carries over the 30 seconds under each scheme. (Under packet switching, traffic that has to wait still gets through a second or two later.)
    Circuit-switched: Packet-switched:

Discuss with your partner

Which type of network would you choose for a university campus network, and why? What usage scenario would benefit from the other type of network?

Q2Transmission Delay vs. Propagation Delay Slides 47–52

For these questions, every message is L = 1,500 bytes = 12,000 bits and:

Transmission delay = L / R, where R is the link rate.
Propagation delay = distance / s.
s = 2×105 km/s through fiber and copper.
s = 3×105 km/s through atmosphere and space.

  1. Scenarios A and B send the message over the same 50 km fiber between Foggy Bottom and the Ashburn campus (VSTC). Only the equipment at the two ends differs. Fill in the table in microseconds.
    Scenario Link rate Transmission delay (µs) Propagation delay (µs) Which dominates?
    A1 Gbps
    B10 Mbps

  2. The link rate is the only thing that changed between A and B. Why did that change one delay term but leave the other exactly where it was?

  3. A geostationary satellite link runs at 100 Mbps and is 36,000 km from the user. Nearly the whole path is through space. Compute both delays in milliseconds, then find the total one-way delay.
    Propagation: ms Transmission: ms Total: ms
  4. Your provider offers to double the satellite link to 200 Mbps. What is the new total one-way delay, and what percentage improvement is that over your answer to Q3?
    New total: ms Improvement: %

Discuss with your partner

Was that upgrade worth buying? Name one application that would clearly benefit from it and one that would barely notice, and say what distinguishes them.

Q3Store-and-Forward and Pipelining Slides 47–58

Names:

A 12,000-bit message travels from source to destination across three links:

Source link 1→ Router 1 link 2→ Router 2 link 3→ Destination

Each link runs at R = 10 Mbps. Propagation, processing, and queuing delays are all zero — transmission delay is the only thing that matters here. Under store-and-forward, a router must receive a packet in full before it can start sending it onward.

Sent as one 12,000-bit packet, the message takes 3 × (12,000 bits / 10 Mbps) = 3.6 ms, because each link sits idle until the one before it has finished.

  1. Now split the message into 3 packets of 4,000 bits, and assume for the moment that the receiver can reassemble them with no extra bits. Each hop now takes ms, which we will call a time slot. Mark which packet (P1, P2, P3) occupies each link during each time slot:
    Slot 1Slot 2Slot 3Slot 4Slot 5Slot 6
    Link 1
    Link 2
    Link 3
    Total time until the last bit arrives: ms

  2. The same 12,000 bits crossed every link in both cases, so why is this faster?

  3. Formulate an equation to generalize from your timeline: sending N packets across M links, where each hop takes t, finishes in × t.
  4. In reality, packets need headers. Suppose each packet now has a 320-bit header carrying addressing and ordering information, on top of its payload. Recompute:
    Split into Payload per packet Total time (ms) Header overhead (%)
    3 packets4,000 bits
    100 packets120 bits

Discuss with your partner

You now have three data points — 1 packet: 3.6 ms, 3 packets: ?, 100 packets: ?. Is smaller always better? What is competing with what, and what does that imply about how big a real packet should be?

Q4Throughput and Bottleneck Links Slides 47–58

A student in a residence hall downloads a large file from a course server.

Server 200 Mbps→ Campus backbone 1 Gbps→ Dorm WiFi 100 Mbps→ Laptop

  1. Suppose only the Campus Backbone network is busy, with 20 users evenly sharing the bandwidth, while the first and third links are idle. Where is the bottleneck and what is the end-to-end throughput?

  2. At 3 a.m. only 2 students are downloading instead of 20. Everything else is unchanged. Recompute the throughput and note which segment limits it now.
    Throughput: Mbps Bottleneck link:

Discuss with your partner

The university has the budget to double exactly one segment: the server uplink, the campus uplink, or the dorm WiFi. Which do you recommend, and what did you have to assume about when students are downloading to justify it? What would be the least useful to upgrade?