Problem
   Implement the CSMA protocol for packet communication between a number of nodes connected to a common bus.

Concepts
   Listen-before-transmit to improve efficiency
   Effect of propagation delay

Program Development
   In ALOHA, a node transmits whenever it wishes without regard to the activities of other nodes. This may cause a collision with the transmission of another node, resulting in both packets being lost.

Under heavy load, frequent collisions can severely reduce the throughput. By adding a modest amount of hardware to the NIU, it is possible to detect the presence of a transmission on the network. We can then modify the transmission protocol to require a node to listen to the network before transmission. Only if the network is idle is the node permitted to transmit. This is referred to as Carrier Sense Multiple Access (CSMA) for historical reasons: it was first used in radio networks where a packet is modulated onto a carrier frequency. The presence of this carrier was sensed.

It is still possible to get collisions due to the non-zero delay between sensing idle and actually starting transmission and the propagation delay of the signal along the bus to other stations. Refer to Figure 1 which shows activity on the network over a period of time. The X-axis represents time, while the Y-axis represents space. Assume that a node A decides to transmit a packet. At time t₀ it senses the network and finds it idle. After some hardware/software latency of δ , it actually starts transmitting at time t₁. This signal propagates at a high but finite speed, and reaches node B at time t₂. Note that t₂ - t₁ = t* δ , where t is the propagation time per metre of the electrical signal (approximately 5ms/m) and d is the distance between nodes A and B.

Thus, even after A has started transmitting, B could sense the cable as being idle and start transmission at time t₃ < t₂. As shown in the figure, B's packet propagates towards A and at some point both packets collide and are lost. Notice that t₃ could even be earlier than t₁. In fact, t₃ could be anywhere in the range (t₁- τ - δ, t₁+ τ + δ ) and cause a collision (see Figure 2). This interval is referred to as the vulnerable period. If B started transmitting before t₁- τ - δ , A would find the network busy at t₁ and would not transmit, avoiding a collision.

First, set the bit delay in the Network Emulator Unit to 0. Measure throughput X for various values of offered load G. Now, increase the bit delay in steps up to the maximum. In each case, measure X versus G. Plot all the curves on a single graph and compare them. Note that the parameter a = (end-to-end delay in bits)/ (packet length in bits) = (bit delay * N)/P, where N is the number of nodes and P is the packet length in bits. Your curves should resemble Figure 3.

Applications
   Especially useful in radio-based LANs

Exercises
   Set up one node as the controller to automate the running of a series of experiments (see Problems under Experiment E4.1).

   Implement non-persistent CSMA. Compare performance to that of 1-persistent CSMA and to the theoretical curves [Tanenbaum 88].

   Implement p-persistent CSMA. Experiment with different values of p. Can you devise an adaptive p-persistent CSMA in which p varies with load and which performs better than with fixed p? Compare performance to that of non- and 1-persistent CSMA.

   Repeat the experiments at higher bit rates. Can you explain any change in the curves?

For more information on what different areas of computer networks and data communication technologies are possible to be covered through Benchmark LanT please visit our website www.benchmark-electronics.com.

| Page 1 |Page 2 | Page 3 | Page 4 | Top |