Enhancement of Token Ring LAN Performance Using Multiple Tokens Technique

Token ring LAN’s have been used for many years and occupied an important part of the computer network world. It has the advantage of stabilized performance over variable load conditions. In this paper, a new method is introduced to enhance the characteristics of the LAN. The idea behind the new method is to give the stations on the ring more opportunity to transmit by splitting the main ring into mutiple subrings each one of them use the same protocol of the ordinary token ring LAN and managed by supervisor stations. Discrete event simulation is used to examine the effect of adopting the new method on the performance of the ordinary LAN(for simplicity of simulation, two subrings is considered in this research). The results obtained from the simulation program show the benefit obtained from the new method.

With the assumption that the number of packets arrived to the channel follow Poisson distribution, the probability of (k) arrivals is given by: P k = ((T/ ) k / k!).e -T/ … (5) Where (T/ ) is the packets arrival rate (packet/sec.) and ( ) is the interarrival time.
The probability of no arrivals is given by: Thus, the probability of at least one of (i) stations having ready packet for transmission: The behavior of a group of stations is practically following the classical binomial distribution, which is represented by the general equation: This equation gives the probability of occurrence of event (G) in an (n) trials for any of (i) sources. Thus, (P i ) as illustrated by equation (2) can be obtained simply by replacing (n) by (Q k ) and (G) by (1-e -T/ ) i .
2. The output of a random number generator (RNG) having uniform distribution with a mean value equal to (1) is taken at a given time slot.
3. If the output of the (RNG) lies between (P ti ) and (P ti + 1 ) then there is (i) stations having packets ready to be transmitted, i.e. KG = KG + i. 4. When a successful transmission occurs, the number of stations having finished their packets transmission is increased by one.
5. The number of stations trying to retransmit again in the next cycle (after finishing their first transmission) is calculated using the following eqs.: If the output of another uniform random number generator lies between (q ti ) and (q ti + 1 ), then there is (i) stations trying to retransmit their packets again. 7. The cumulative number of stations wishing to transmit in the next scanning cycle is equal to the initial value of (KG).
From the simulation point of view, there is a free token arrived at the supervisor station only if the number of stations on a subring trying to transmit become zero (Kg = 0). The simulator checks the buffers of the supervisor station in order to discover its entity to transmit. If the buffers are empty (i.e., packets counter = 0), the supervisor station retransmit the free token again. On the other hand, if there are stored packets in the buffers, then the supervisor station transmits all the packets one after the other then updating the packets and buffers counters which are used to calculate the buffer time delay.
The flow chart shown in Fig. (3) illustrates the performance of the modified token ring simulator.
As a matter of comparison, the performance of a conventional and modified token ring LAN's with different conditions are studied using simulation technique. The following parameters and variables are suggested during the run of the simulation process:

4.Bit rate = 4 Mbps.
Figures (4 & 5) shows the throughput and the average packet delay as a function of the offered load( ), it is obvious that at low offered load, there are small numbers of packets ready to be transmitted, hence, small values of throughput and average packet delay are expected. As the offered load increases, more packets are arrived to the channel, and the busy cycles of the channel increase too, causing an increment in the throughput and a longer average packet delay (a station has to wait longer time before getting a free token ). The deterministic nature of the token ring medium access protocol causes the throughput and the average packet delay to be constant at very high offered load (in such a case, all stations have ready packets for transmission). The figures also indicate that the modified token ring LAN has better performance as compared with the conventional one (especially at high offered load conditions).
The effect of load distribution (the percentage of the load from a subring directed to the other subring) in the modified token ring LAN is studied. Better performance could be obtained if the load circulating in a subring is much greater than the transferred load to the other subring. As a consequence, an increment in the total throughput could be measured due to the simultaneous transmissions on the two subrings and a decrement in the average packet delay due to the shorter path of transmission (on a single subring) as compared with the path around the original token ring LAN.
Confidence about the correctness of the simulation results is ensured, this is achieved by comparing the performance results of a single subring using the modified token ring simulator with that of the ordinary (conventional) token ring simulator on the condition that the working parameters are the same. The following parameters are set to both simulators: