Performance Analysis of Transmission Control Protocol Variants

Transcription

1 IJCST Vo l. 3, Is s u e 1, Ja n. - Ma r c h 212 ISSN : (Online) ISSN : (Print) Performance Analysis of Transmission Control Protocol Variants 1 Pooja Sahni, 2 Dr. Ravinder Khanna 1 Research Scholar, School of Electronics and Electrical Engineering, Singhania University, Pacheri Bari, Rajasthan, India 2 Sachdeva College of Engg. for Girls, Kharar, Mohali, Punjab, India Abstract With the emergence of new technologies over the last decade, the uses of handheld devices and laptops have become very common. Most of these devices interconnect using wireless links. Transmission Control Protocol (TCP) is the most widely accepted, reliable, connection-oriented, full-duplex, byte stream transport level protocol that is in use today. In wired networks, whenever packet loss occurs, it is mainly due to the congestion, whereas in case of wireless, it may be because of atmospheric noise, lightening, link breakage and other noise signals in addition to congestion. It is important to improve its performance in wireless networks without any modification to the application interface provided by TCP on fixed hosts. This is the only way by which mobile devices communicating on wireless links can seamlessly integrate with the rest of the Internet. The simulator we used for comparison is Network Simulator (NS2).In this paper we will discuss the performance of various TCP Protocols Keywords TCP, TCP Reno, Time Delay, Throughput, Wireless Network I. Introduction The Transmission Control Protocol (TCP) is used for a highly reliable host-to-host protocol between hosts in packet-switched computer communication networks, and in interconnected systems of such networks. TCP is a connection-oriented, end-to-end reliable protocol designed to fit into a layered hierarchy of protocols which support multi-network applications. The TCP provides for reliable inter-process communication between pairs of processes in host computers attached to distinct but interconnected [1], computer communication networks. The exponential rise in the internet users over the last few years has created congestion problem over the Wide Area Network (WAN) and has become more and more severe with the exponential rise in the internet users. So, to avoid congestion problem, there is a need of some congestion avoidance protocols. Over the internet, TCP/IP is generally used protocol. As the global Internet traffic increases, many popular sites are often unable to serve their TCP/IP workload, particularly during peak periods of activity [2]. A network that uses internet technology is called an internetworking [3]. TCP/IP model is made up of 4 layers i.e. Application layer, Transport layer, Internet layer, Network layer.transmission Control Protocol (TCP) is one of the main protocols used at the transport layer.the TCP/IP model is shown as below. II. Congestion Avoidance Congestion is the phenomenon that occurs at a router when incoming packets arrive at a rate faster than the router can switch (or forward) them to an outgoing link. However, it is important to distinguish contention and congestion. Contention occurs when multiple packets have to be queued at a switch (or a router) because they are competing for the same output link, whereas congestion means that the switch has so many packets queued that it runs out of buffer space and has to start dropping packets. Congestive collapse (or congestion collapse) is a condition which a packet switched computer network can reach, when little or no useful communication is happening due to congestion. Congestion collapse generally occurs at choke points in the network, where the total incoming traffic to a node exceeds the outgoing bandwidth. A. Classification of Congestion Control Algorithms There are many ways to classify congestion control algorithms: 1. Slow Start Algorithm The Slow Start Algorithm tries to avoid congestion by sending data packets defensively. Therefore, two special variables named congestion window (cwnd) and Slow Start threshold (ssthresh) are stored on sender s side. Initially, cwnd is sized to one packet when the sender injects a new packet into the network and waits for the Acknowledgment (ACK) [4]. from the receiver. Normally, this packet gets through the network and reaches the recipient in time, so it will be replied by an ACK. If this acknowledgment is received by the sender, cwnd is incremented; if network capacity is reached and packets get lost, the sender does not increment the number of packets any further. That means, by each sending cycle the number of injected data packets is doubled until network s capacity is reached and the required ACK cannot get through. 2. Fast Retransmit Algorithm Fast Retransmit Algorithm uses explicit feedback methods to avoid long timeout periods waiting for packet retransmitting in case of packet loss. Such problems are inherent in packet-switched data networks because every data packet can travel individually through the rest of the network and can use special routes from the sender to the recipient [5]. Consequently, the transmitted data packets will neither reach the recipient in accurate order nor complete continually. Therefore, after detecting a missing packet the recipient sends duplicated ACK packets for the last correct received packet until the missing packet receives. Unfortunately, TCP may use duplicate ACK packets to indicate out-of-orderpackets, thus two ACK packets do not necessarily indicate a lost packet. Therefore, if a sender receives multiple ACK packets with the same sequence number, normally at least three of them, these packets indicate the last successfully transmitted packet. Fig. 1: TCP/IP Layers Model 382 In t e r n a t i o n a l Jo u r n a l o f Co m p u t e r Sc i e n c e An d Te c h n o l o g y

2 ISSN : (Online) ISSN : (Print) 3. Fast Recovery Algorithm A special Congestion Avoidance Algorithm often combined with Fast Retransmit to restart transmission at a higher throughput rate than Slow Start is the FAST RECOVERY Algorithm. Fast Recovery starts when Fast Retransmit fails to work. If no further duplicate ACK packets are received for Fast Retransmit Algorithm the sender tries to return to normal sending state. III. Classification of TCP Protocols A. Packet Loss-based TCP Protocols Variants These are the protocols which uses packet drop probability as the main factor for adjusting the window size. These variants of TCP use congestion control algorithms. There were developed initially and are still used. Loss based TCP protocols are more aggressive than the delay based TCP protocols [6]. These are classified as below. Congestion has abrupt change in window size Recovery is slow. IJCST Vo l. 3, Is s u e 1, Ja n. - Ma r c h 212 B. Delay-Based TCP Protocols Delay-based algorithms were developed so as to provide stable throughput at the receiver end. These TCP variants use congestion avoidance algorithms to avoid the packet loss and are less aggressive than packet loss based TCP protocols. Delay-based algorithms can maintain a constant window size, avoiding the oscillations inherent in loss-based algorithms [8].These are classified as below. 1. TCP Vegas TCP Vegas is a congestion control or network congestion avoidance algorithm that emphasizes packet delay, rather than packet loss, to determine the rate at which to send packets. TCP Vegas detects congestion during every stage based onincreasing Round Trip Time (RTT) values of the packets in the connection unlike Reno, Tahoe etc. which detect congestion only after it has actually happened via packet drops [9]. TCP Vegas adjusts the source rate before actually packet is dropped. Queuing delay is the difference between basertt and avgrtt. TCP Vegas decreases the source rate in case of increase in queuing delay value and increases in case of decrease in queuing Fig. 2: Classifications of TCP Protocols 1. TCP Tahoe It uses Additive Increase Multiplicative Decrease (AIMD) algorithm to adjust window size. It means that increases the congestion window by one for successful packet delivery and reduces the window to half of its actual size in case of data loss or any delay only when it receives the first negative acknowledge. During Slow Start stage, TCP Tahoe increases window size exponentially i.e. for every acknowledgement received, it sends two packets. During Congestion Avoidance, it increases the window size by one packet per Round Trip Time (RTT) so as to avoid congestion. In case of packet loss, it reduces the window size to one and enters in Slow Start stage. 2. TCP Reno TCP Reno is the modified variant of TCP Tahoe, suggested by Jacobson in 199. TCP Reno works very much similar to TCP- Tahoe. Packet loss detected via duplicate ACKs results in the window being cut by half [7]. If a timer expiry does occur, then the window size is dropped to one, and slow start is used to grow the window back to half its value when the timer expired. During the transmission, if three duplicate ACKs are received, Reno will halve the congestion window, perform a fast retransmit, and enter a phase called Fast Recovery. If an ACK times out, slow start is used as it is with Tahoe. Drawbacks of TCP Reno and TCP Tahoe: Low performance on fast long distance networks IV. Performance Analysis Network protocols are studied by simulating the network scenarios using software like NS2. Ns-2 is a discrete event network simulator which simulates events as sending,receiving, dropping and forwarding packets etc. A. Topology Used for Simulation For the comparison of different TCP variants, simulation is done for the dumbbell topology as shown in fig. 3, in which there are three source nodes (i.e. S1, S2 and S3) which are sending data to sink nodes (i.e. D1, D2 and D2)[1] through a bottleneck link between nodes S and D. Steps used in writing a simulating script are: Create the event scheduler Turn on tracing Create network Setup routing Insert errors Create transport connection Create traffic Transmit application-level data D acts as router which forward data to the sink nodes over the network. The delay for all the side links is kept constant as shown in fig. 3. Simulation is done for two values of link capacities (C) i.e. for C = 1 Mbps and C = 1 Gbps. Delay on the bottleneck link (i.e. X) is varied on bottleneck link and simulation is done for four values of X i.e. for X=8 ms, 18 ms, 48 ms and 98 ms, so as to make total delay from source node to the sink node equals to 1ms, 2ms, 5ms and 1ms respectively. Simulation is done for 1 seconds in every case and window size is kept very high (i.e. cwnd=1) in all the cases so that maximum bandwidth may be used. International Journal of Computer Science And Technology 383

3 IJCST Vo l. 3, Is s u e 1, Ja n. - Ma r c h 212 ISSN : (Online) ISSN : (Print) At 2ms Delay Fig. 3: Dumbbell Topology 1. TCP Tahoe s Throughput Variation with Time for Delay 1ms 2ms 5ms AT 1 Mbps Link Capacity At 1ms Delay Fig. 4: At 2ms Delay At 5ms Delay TCP Tahoe at 1 Gbps :Throughput Variation with Time for Delay. 1ms. 2ms. 5ms and 1 Gbps link Capacity At 1ms Delay At 5ms Delay Fig. 5: From the graphs for TCP Tahoe we can take two cases: When link capacity of all the channels is 1 Mbps. As the delay the throughput also decreases. TCP Tahoe attains maximum throughput for 1ms time delay and falls significantly to a much smaller value for 1ms time delay and the bandwidth utilization is 41.8% at 1 ms delay and decreases to 9.16% at 1 ms time the throughput also decreases. TCP Tahoe attains maximum throughput for 1ms time delay and falls significantly to a much smaller value for 1 ms time delay but the bandwidth utilization is 24.1% for 1ms and decreases to.92% at 1ms time 3. TCP Reno s Throughput Variation with Time for Delay 1ms. 2ms. 5ms and 1 Mbps Link Capacity At 1ms Delay In t e r n a t i o n a l Jo u r n a l o f Co m p u t e r Sc i e n c e An d Te c h n o l o g y

4 ISSN : (Online) ISSN : (Print) Fig. 6: At 2ms Delay At 5ms Delay TCP Reno s Throughput Variation with Time for Delay 1ms. 2ms. 5ms and 1 Gbps Link Capacity At 1ms Delay At 2ms Delay At 5ms Delay IJCST Vo l. 3, Is s u e 1, Ja n. - Ma r c h 212 From the graphs for TCP Reno we can take two cases : When link capacity of all the channels is 1 Mbps. As the delay the throughput also decreases. TCP Reno attains maximum throughput for 1ms time delay and falls significantly to a much smaller value for 1ms time delay and the bandwidth utilization is 43.24% at 1 ms delay and decreases to 9.35% at 1 ms time the throughput also decreases. TCP Reno attains maximum throughput for 1ms time delay and falls significantly to a much smaller value for 1 ms time delay but the bandwidth utilization is 29.1% for 1ms and decreases to.94% at 1ms time 4. TCP Vegas s Throughput Variation with Time for Delay 1ms. 2ms. 5ms and 1 Mbps Link Capacity Fig. 8: At 2ms Delay At 5ms Delay TCP Vegas s Throughput Variation with Time for Delay. 1ms. 2ms. 5ms and 1 Gbps Link Capacity At 1ms Delay Fig. 7: International Journal of Computer Science And Technology 385

5 IJCST Vo l. 3, Is s u e 1, Ja n. - Ma r c h At 2ms Delay At 5ms Delay Fig. 9: From the graphs for TCP Vegas we can take two cases: When link capacity of all the channels is 1 Mbps. As the delay the throughput also decreases. TCP Vegas attains maximum throughput for 1ms time delay and falls significantly to a much smaller value for 1ms time delay and the bandwidth utilization is 53.8% at 1 ms delay and decreases to 7.31% at 1 ms time the throughput also decreases. TCP Vegas attains maximum throughput for 1ms time delay and falls significantly to a much smaller value for 1 ms time delay but the bandwidth utilization is 4.5% for 1ms and decreases to.73% at 1ms time Collectively we can say that TCP Reno s throughput is slightly better than TCP Tahoe. TCP Vegas provides best throughput than TCP Reno and TCP Tahoe. As the delay increases, the throughput falls significantly for all the three protocols. Also From the graphs, it is concluded that TCP Reno and TCP Tahoe provides oscillatory throughput i.e. throughput keep varying significantly with time. V. Conclusions TCP Vegas is better than any other TCP variants for sending data and information due to its better Packet Delivery Fraction and Avg. end- to- end delay in both high and low mobility. It provides best throughput than TCP Reno and TCP Tahoe. This is due to fine tuning of congestion window size by taking into consideration the RTT of a packet, whereas other reactive protocols like TCP Tahoe, RENO, New RENO, SACK, and FACK continue to increase their window size until packet loss is detected. VI. References [1] C. Jin, Fast TCP: From Theory to Experiments, IEEE Network, Vol. 19, No. 1, pp. 4-11, 25. ISSN : (Online) ISSN : (Print) [2] Fei Ge, Liansheng Tan, Moshe Zukerman, Throughput of Fast TCP in Asymmetric Networks, IEEE communications letters, Vol. 12, No. 2, 28. [3] Kyungmo Koo, Joon-Young Choi, Jin S. Lee, Parameter Conditions for Global Stability of fast TCP, IEEE communications letters, Vol. 12, No. 2, 28. [4] J. Wang, D. X. Wei, S. H. Low, Modelling and Stability of Fast TCP, in proc. IEEE infocom 25, miami, fl, March 25. [5] J. Wang, A. Tang, S. H. Low, Local stability of Fast TCP, proc. IEEE conf. decision and control, December 24. [6] C. Jin, D. Wei, S. H. Low, Fast TCP for high-speed longdistance networks, Internet draft draft-jwl-tcp-fast-1.txt. [7] David X., Wei Cheng Jin, Steven H. Low Sanjay Hegde, Fast TCP: Motivation, Architecture, Algorithms, Performance, IEEE/ACM Transactions on Networking, 14(6), pp , 26. [8] T.V. Lakshman, Upamanyu Madhow, Bernhard Suter, TCP/IP Performance with Random Loss and Bidirectional Congestion, IEEE/ACM Transactions on networking, Vol. 8, NO. 5, 2. [9] Cheng Peng Fu, Bin Zhou, Jian Ling Zhang, Modeling TCP Veno Throughput over Wired/Wireless Networks, IEEE communications letters, Vol. 11, No. 9, 27. [1] Cheng P. Fu, South C. Liew, A Remedy for Performance Degradation of TCP Vegas in Asymmetric Networks, IEEE Communications letters, Vol. 7, No. 1, 23. Born in 1982, Er. Pooja Sahni Graduated in Electronics & Communication Engineering from Kurukshetra University in 23. She completed her Masters in Electronics & Communication Engineering Engineering from Kurukshetra University in 29. Presently she is perusing her Ph.D in Electronics & Communication Engineering. She worked as a senior lecturer for three years. Currently, she is working as an Asstt. Professor in Indoglobal College of Engg. She has a total experience of 9 years in teaching. Her area of interest are Digital Communication, Computer Network and optical communication. Born in 1948, Dr. Ravinder Khanna Graduated in Electrical Engineering from Indian Institute of Technology (IIT) Dehli in 197 and Completed his Masters and Ph.D degree in Eletronics and Communications Engineering from the same Institute in 1981 and 199 respectively. He worked as an Electronics Engineer in Indian Defense Forces for 24 Years where he was involved in teaching, research and project magement of some of the high tech weapon systems. Since 1996 he has full time Switched to academics. he has worked in many premiere technical institute in india and abroad. Currently he is the Principal of Sachdeva Engineering College for Girls, Mohali, Punjab (India). He is active in the general area of Computer Networks, Image Processing and Natural Language Processing. 386 In t e r n a t i o n a l Jo u r n a l o f Co m p u t e r Sc i e n c e An d Te c h n o l o g y