Advanced TCP/IP networks midterm study guide    Midterm: Wed Mar 13, 2002

REVISED VERSION -- this version is more tuned to Spring 2002 coverage.

See the notes file for the topics covered. Essentially we've looked at:
        
        IP routing issues (eg provider-based routing)
        
        TCP:
                self-clocking
                sawtooth form
                avg_cwnd = k/sqrt(loss_probability), k = constant
                Basic algorithms: tahoe, reno, newreno, SACK
                Fairness issues
                
                
Some specific sections of Peterson & Davie:

Chapter 4:
    4.2.2 (and some of 4.2.3): basics of IP routing, 
           including Distance-Vector algorithm
    4.3.2: CIDR
    
Chapter 6:
    6.1:   Good overview of some issues
    6.2.1: FIFO == droptail
    6.3:   TCP mechanisms, Tahoe & Reno
    6.4:   DecBit, RED, packetpair (we discussed packetpair, not the others)
    6.5.1: playback buffers, constant-bit-rate (CBR) traffic, jitter
    
=============================================================
    
Some sample questions:

Chapter 4 exercises that are worth at least reading:
17, 18, 24, 25, 26, 38, 39, 40, 41

Chapter 6 exercises that are worth at least reading:
5, 6, 11, 15, 24, 25, 27, 28, 39a

1. Give a timeline showing two packet drops in one TCP window, 
for both Reno and NewReno implementations. Identify periods of 
congestion window "inflation" (the upper end moves up but the 
lower end does not) and "deflation" (the lower end moves up but 
the upper end does not).

2. How do Reno and NewReno handle "partial ACKs"? That is,
suppose packet1 is lost, with a window size of 20. Both TCPs
would receive duplicate ACK0's, and would retransmit packet1.
If packet1 is the only packet lost, the response to the
retransmitted packet1 will be ACK20. However, if, say, packet12
were also lost, then the response to the retransmitted packet1 
would be ACK11. This ACK11 represents a "partial ACK".
How does SACK TCP respond to partial ACKs?

3. Explain the connection between cwnd and estFlightSize, 
both at loss-free times and in the presence of packet losses.

4. Explain why a longer RTT alone, with the same packet-loss risk, 
can mean that a TCP connection gets less bandwidth.

5. Discuss how to arrange for geographical routing in IP.
Discuss how routers in adjacent towns on opposite sides of a state 
line could send traffic to each other directly, while each sends 
other traffic to their state's respective state hub.

6. Suppose A sends to B via some routers and uses the sliding-window
algorithm with a fixed window size W; the RTT is 2.0 sec of
bandwidth-type delay (mostly for the A->B direction; ACKS, being
smaller, are delayed much less). Suppose the bandwidth is 2 packets/sec.
Describe what happens as W increases from 1. At what value of W
would an increased RTT be encountered? What causes this? 
What happens if W is increased just a little more?
When might the sender encounter lost packets?

7. (a). Explain why BGP speakers exchange full AS-paths to each 
destination.
(b). Explain the tradeoff between route aggregation using CIDR
(eg of all routes to a provider's customers and subcustomers
into a single entry), and the use of AS-paths as above.
How is this usually resolved?