International Journal of Computer Applications (0975 – 8887) Volume 16– No.2, February 2011 40 Fairness and Percentage Loss Analysis for Short Lived Web Transfers using QMBCCA ABSTRACT One of the major problems in Internet today is Network Congestion. TCP is used as a Transport Protocol for most of the applications on the Internet and is also responsible for adjusting to network congestion dynamically. The Active Queue Management algorithms (AQM) are used to reduce congestion, and in this paper two such AQM algorithms are considered. Here we look at the existing Explicit Congestion Notification (ECN) algorithm and a Queue Management Backward Congestion Control Algorithm (QMBCCA). A comparison is made between the two algorithms in terms of Fairness and Percentage Loss for short lived web transfers. We have found that there is a significant reduction in packet loss and improvement in fairness index when using QMBCCA when there were more number of TCP flows. Key Words Congestion Control, Active Queue Management, RED, ECN, Packet Loss, Fairness 1. INTRODUCTION Congestion occurs when multiple input streams arrive at a router whose output capacity is less than the sum of the inputs [1]. The normal method that a router reacts to congestion is by dropping the incoming packets in the absence of buffer space. This is commonly called as “Tail Drop” or “Drop Tail”. The dropping of incoming packets is an indication to the transport layer that there is congestion in the network. TCP is the most leading transport protocol today [2]. Whenever congestion occurs in the network, an intermediate router begins to drop packets. In response to this hint of congestion, TCP invokes its congestion control algorithms [3] so that congestion is eased. These congestion control algorithms make TCP responsive even to a single packet loss. The use of more web applications which use TCP as their transport protocol has resulted in an increase in network traffic, which leads to, increased network congestion and increased packet drops. This affects the performance of TCP drastically. With the increase in the amount of traffic and congestion in the Internet, it is no longer practical to rely only on the congestion control algorithms to provide good service at al1 times. Many research works have gone into how the network can involve itself to complement these congestion control mechanisms. This is where the Queue Management algorithms are used “Queue Management” algorithms deal with managing the lengths of packet queues by dropping packets whenever necessary. The main concept of these algorithms is to involve the routers for easing congestion before it actually occurs in the network. Dropping packets probabilistically even before the queue fills up in the router eases congestion. This makes the sender TCP to back off and ease congestion before it actually occurs. Random Early Detection (RED) is one of the queue related algorithms that is popular in recent years. RED [4] follows the concept of Early Random Drop by introducing the measure of average queue size and also dynamically changing the drop probability. The algorithm, called Loss Ratio-based RED (LRED), measures the latest packet loss ratio, and uses it as a complement to queue length for adaptively adjusting the packet drop probability. This achieves fast response time and yet good robustness.[5] Explicit Congestion Notification (ECN) algorithm is an extension of the RED algorithm that marks a packet instead of dropping it. [6] proposes a new AQM algorithm that exploits Round Trip Time explicitly by introducing a passive RTT estimation technique in a router. PERED (pre-Estimation RED) is a new active queue management algorithm which creates a dynamical Markov model according to the traffic loads to estimate the future queue size of router buffer and self-adjusts the parameters of well-known Random Early Detection (RED) to get better performance[7]. RED requires careful tuning of its parameters in order to provide good performance. A new mechanism BO-ARED (An Adaptive RED Algorithm Combined With Buffer Occupation) improves the performance of RED by matching router's buffer occupation with w q , min th , max th and P max parameter settings, to make BO-ARED adapt to network environment variation automatically[8]. The QMBCCA algorithm is the algorithm that has been proposed by us [9]. QMBCCA work in combination with RED and ECN to provide congestion feedback to the sender without actually dropping the packets. The advantage is that it reduces the number of packet drops, which is expected to benefit TCP during times of congestion. The paper is organised as follows. Section 2 gives an introduction on the Active Queue Management (AQM) Algorithms, specifically explaining about the RED and ECN Algorithms. Section 3 explains the QMBCCA algorithm. The Simulation Screenshots and the corresponding graphs are given for the two parameters chosen for comparing the two algorithms in Section 4. Section 5 concludes the paper. 2. AQM ALGORITHMS Active Queue Management algorithms basically detect congestion before the queue overflows and it also gives an indication of the congestion that has occurred to the end nodes [4]. The transport protocols do not have to rely on buffer overflow as the only indication of congestion in an active queue management. This is because they get notified before congestion happens. The active queue management allows routers to control and count how many Dr.E.George Dharma PrakashRaj Asst Professor in Computer Science Bharathidasan University Tiruchirappalli, India. Dr. D.I.George Amalarethinam Director - MCA Jamal Mohammed College Tiruchirappalli, India. V. Sinthu Janita Asst Professor & Head Dept of Computer Science Cauvery College for Women Tiruchirappalli, India.