In this lesson, we focus on CSMA and the CSMA with Collision Detection for Random Access Control. The low maximum Throughput of Aloha protocol is due to the wastage of transmission bandwidth because of frequent collisions. When collisions occur they involve entire frame transmission time. These collisions can be reduced by avoiding transmissions that are certain to cause collisions, by sensing the communication medium for the presence of a carrier signal from other stations. A station can determine whether there is an ongoing transmission. This capability is called Carrier sense. In Carrier Sensing Multiple Access, CSMA, schemes a station sense the channel before it starts transmission. If busy, it either wait or backoff. If idle, it starts transmission. From the two station example, we learn that the station would capture the channel after one propagation delay. Therefore, vulnerable period is reduced to one propagation delay. There are different CSMA options, mainly due to the backoff strategies. One persistent CSMA is most aggressive. It starts as soon as the channel becomes idle, the benefits are low latency. But the problem is a low efficiency due to relatively high collision rate. The Normalized propagation delay alpha has a significant impact on the maximum achievable Throughput of a CSMA scheme. Alpha is a ratio of propagation delay over the frame transmission time. In one persistent CSMA, its maximum Throughput throughput is better than that of Aloha schemes for small alpha. But it is worse as alpha is greater than one. The other two CSMA options are non-persistent CSMA and p-persistent CSMA. Non-persistent CSMA is the least greedy. If the channel is busy, it immediately run for a backoff period and sense carrier again. If the channel is idle, the station transmits. The advantage is high efficiency but the constraint is longer delay. The p-persistent CSMA is adjustable greedy. It combines elements of one persistent and non-persistent CSMA options. If the channel is busy, a station persists with sensing until the channel becomes idle. If the channel is idle with probability p, a station transmits. With probability 1 minus P, it waits an additional propagation delay before sensing the channel again. It balances delay and efficiency. For non-persistent CSMA, it achieves a higher maximum Throughout than one persistent CSMA, for small Normalized propagation delay alpha but it is even worse than Aloha when alpha is greater than 1. CSMA schemes improve over Aloha scheme by reducing the vulnerable period from one or two frame transmission times to just a single propagation delay. However, collisions still involved in higher frame transmission time. If a station can detect that a collision is taking place, it can abort the transmission right away so wastage would be reduced to time to detect collision and abort a transmission. Quickly or terminating a damaged frame saves time and bandwidth because generally frame transmission time is far greater than propagation delay. This is called a CSMA with collision detection. If Collisions are detected or if stations involved abort their transmissions reschedule a random backoff time and try again at a scheduled times, note here the station can choose any of the CSMA backoff options. From the two station examples in the previous lesson, we learned that it takes two propagation times to find out where the channel has been captured. This is the action time of CSMA with collision detection. In CSMA with collision detection model, collisions can be detected and resolved into two propagation times, each contention slot is therefore two propagation delays. Once the contention period is over, a station takes frame transmission time. It takes one propagation for a station to find out that a transmission is just over. And then to start a contention again. To know the efficiency of the CSMA with collision detection model, we needed to know how long does it take to resolve contention on a successfully captured channel? We know that contention is resolved if exactly only one station transmits in a contention slot. Let's say there are n active stations in a network and each station may transmit with probability p in each contention time slot. We have the probability of successful capture equal to n times p times 1 minus P plus n minus 1. But, is one station transmits in probability p and the other n minus one stations, each is not transmitting with probability 1 minus P. By taking derivative of P success, we find the maximum occurs when P equals to 1 over n therefore on average it takes e time to resolve contention. Referring to the CSMA collision with detection model for a successful frame transmission, there is a need of e propagation delay plus one propagation delay for the current transmission to be successful, so the maximum efficiency is derived. As a summary, we compare the maximum slope Throughput for Random Access MAC's protocols. When the normalized propagation delay factor alpha is small, CSMA with collision detection has the best maximum Throughput but when alpha is large, Aloha schemes have better maximum Throughput since doesn't depend on Alpha factor.
