Computer Networks

Computer Networks
Computer networks like internet necessitate network resources, i.e. bandwidth, buffer spaces, etc in order to accommodate the arriving packets at router buffers [25]. When the arriving packets cannot be accommodated due to lack of network resources, this indicates occurring congestion at router buffers of networks [21]. Congestion can deteriorate a network performance [22, 25] through growing the packet loss probability due to overflow as well as the mean waiting time for packets in the queueing network. In addition, congestion may reduce the throughput and increase the packet dropping probability precedes the router buffers have overflowed.
Congestion can also generate unmaintained average queue length ( ), and this may lead to build up the contents of router buffers, and thus many arriving packets may drop or lose at the router buffers.
Many researchers have proposed methods with aim to control congestion at router buffers of networks [1, 2, 4, 5, 8, 14, 15, 16], i.e. AQM methods [1, 2, 4, 5, 6, 7, 8, 13, 14, 15, 16]. Each AQM is proposed as a congestion control method that identifies congestion at router buffers in an early stage, which means before the router buffers have overflowed. The most known AQM method is RED [14], Gentle RED [16], the Adaptive GRED [2], Adaptive RED [15], Random Early Marking (REM) [7], Dynamic Random Early Drop (DRED) [8] and some discrete-time queue analytical models [1, 4, 5, 6] which were constructed depending on some of AQM techniques. For example, DRED analytical models which they constructed by analysing two queue nodes [5] and three queue nodes [1] based on DRED and using discrete-time queues mechanism [26]. GRED analytical model [4] and BLUE analytical model [6] were constructed depending on GRED and BLUE, respectively and utilising discrete-time queues mechanism.
As mentioned previously, RED was proposed as a congestion control method, but RED can degrade the network?s performance due to the following causes: 1) abruptly RED can increase its arrival rate aggressively, thus the RED?s router buffers may overflow. Therefore, every arriving packet will lose. 2) At a particular time, the RED?s congestion measure ( ) value may be below the value of minimum threshold position at the router buffer ( ). This indicates no packet can be dropped. However, for a short time the arrival rate increases and making the router buffer overflowing. Nonetheless, the value increases but still smaller than the value of . This also makes no dropping for packets even the router buffer is overflowing. 3) RED is greatly reliance upon setting of its parameters ( , the maximum position at the router buffer ( ), queue weight ( ), maximum value of packet dropping probability ( )) [14] in order to obtain a satisfactory performance [9, 15].
To deal with the above RED?s problem, GRED was proposed as a variant of RED, in which GRED enhances the way of tuning some of RED?s parameters such as and [16]. Moreover, GRED maintains the value at a level on a router buffer which is between the and positions, this level is named [16]. Maintaining value at level avoids increasing the router buffer size to be above the position, and therefore smaller number of packets is dropped.
Some researchers have proposed fuzzy logic controller (FLC) mechanisms in order to use them as congestion control approaches [10, 11, 12, 17, 27]. A reason for developing these FLC mechanisms is the difficult of developing new AQM mechanisms [20, 27]. [27] proposed a FLC mechanism aims to enhance the BLUE method?s performance [13]. Two input linguistic variables (current queue length and packet loss rate) were utilised to evaluate a single output linguistic variable (packet dropping probability). The performance measure results with reference to queue length, throughput and packet loss rate of the FLC mechanism in [27] were provided better than those of BLUE method. [10, 11, 12, 17] proposed a FLC mechanism based on RED method which applied in TCP/IP differentiated service networks. In [10, 11, 12, 17], different classes of services are performed as well as different linguistic rules for every class of service. The FLC mechanism depends on two input linguistic variables (current queue length, the change rate in the traffic load), which used to produce a single output linguistic variable (packet dropping probability). [10, 11, 12, 17] were arisen that the FLC mechanism outperformed the RED method in terms of to throughput and the queue size results. In [24], another FLC mechanism was proposed and called Adaptive Fuzzy RED (AFRED), this mechanism relies upon a single input linguistic variable (current queue length) in order to compute a single output linguistic variable (packet dropping probability). The experimental results have revealed that AFRED mechanism outperformed RED method regarding to the throughput results and in maintaining the queue size as low as potential.
This paper introduces a comparison between three AQM methods (GRED, REDD1 and the Adaptive GRED) with reference to several performance measures ( , , , , ) with aim to identify which method offers more satisfactory performance measure results in cases of occurring of congestion or not. The reason of choosing GRED and REDD1 methods in the comparison is due to these methods were proposed to enhance the RED?s performance. The Adaptive GRED is chosen since it is based on GRED and was proposed to improve the performance of GRED with respect to , and in presenting of congestion, and also the Adaptive GRED enhances the parameter settings of and .
The structure of this paper is as the following: GRED method is introduced in Section 2. Sections 3 and 4 present the FLC method based on RED called REDD1 and the Adaptive GRED, respectively. A simulation details about GRED, REDD1 and the Adaptive GRED methods are highlighted in Section 5. Section 6 demonstrates the simulation results of the performance measure results for GRED, REDD1 and the Adaptive GRED methods. Lastly, paper?s conclusions and future work are given in Section 7.

Computer Networks 9 of 10 on the basis of 4218 Review.