Unit home

help3002

Lecture recordings
via LMS

Schedule

Project

Final exam

Unit outline

Resources

Textbooks

Extra reading

Look after yourself!

Labsheet 3 - for the week commencing 18th March 2024

1. [1-2 hours] [files required: stopandwait.c, STOPANDWAIT1 ]
   In Lecture 3 we introduced the stop-and-wait Data Link Layer protocol, which delivers frames between directly connected hosts. We also recognized that the nature of stop-and-wait results in a lot of time being 'wasted', particularly in the presence of Physical Layer errors. In Labsheet 2 we observed that employing negative-acknowledgments can reduce the waiting time, resulting in an increased data throughput.
   Another mechanism to improve stop-and-wait is to piggyback acknowledgments onto data frames. Each frame now carries two sequence numbers: a frame travelling from A→B may include a sequence number describing data being carried from A→B and may include another sequence number acknowledging data previously sent from B→A. The benefit is that fewer frames need to traverse 'the wire', and fewer frames interrupt each network interface-card and operating system.
   The implementation of piggybacking is a modification to stop-and-wait:
   • When the receiver gets a DL_DATA frame, it does not immediately send an DL_ACK frame.
   • The receiver waits (for a limited time) until it has its own outgoing DL_DATA frame to send, and piggybacks the pending DL_ACK in the outgoing header.
   • If no DL_DATA frame becomes available in a short time, the receiver must send an DL_ACK frame by itself.

   For this task, using the cnet network simulator:
   • Add piggybacking to the standard stop-and-wait protocol.
   • Modify the topology file, changing the attributes of message delivery, bandwidth, frame corruption and loss, to determine under which conditions piggybacking offers an advantage.

2. To perform a valid scientific experiment (which, after all, is what we're doing), we need to hold all conditions/arrtibutes constant and vary just one over time, perform the same experiment for both the stop-and-wait and piggybacking protocols.
   As in Labsheet 2, develop a plot to help you to visually compare the results of your experiment. The easiest approach is to use the plot-to-html.sh shellscript used in Labsheet 2 to plot multiple curves on the same graph (you may like to learn about the 'join' program to merge the results of the two simulations).

3. [HARDER 1-2 hours, and requires reading a textbook] 🌶 🌶
   Sliding-window protocols are Data Link Layer protocols that support multiple outstanding frames between sender and receiver. Again, the goal is to minimize the waiting time required in the standard stop-and-wait protocol.
   In general, sliding-window protocols have the properties that:
   • the sender has a sending window consisting of a sequence (an array) of frames that have been sent but not acknowledged.
   • The sender's window size grows as more frames are sent but not yet acknowledged.
   • The receiver has a receiving window consisting of frames it is willing to accept. The receiver's window size remains constant.
   • Each frame being sent has a sequence number from 0 to 2^n-1 (which fits in >n bits). Stop-and-wait has n=1.
   • A window size of 1 implies that frames are only accepted in order, and thus an error is detected if an expected frame is missed (a 'later' frame arrives first).
   The simplest variant of a sliding-window protocol is termed the go-back-N protocol, in which the receiver simply discards (ignores) all frames after a bad one. This corresponds to a receiver window size of 1.
   After it learns of a lost (or corrupted frame), the sender retransmits all unacknowledged frames - it "goes back N frames" and starts again. Notice that in the presence of significant errors that bandwidth is wasted because the receiver only 'buffers' a single frame.
   Following the same experimental approach as in the first task, use cnet to implement the go-back-N protocol, and determine under which conditions it is an improvement over the stop-and-wait protocol.