Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                

TCP-Performance for Wireless Environment

2018

Improving the performance of the transmission control protocol (TCP) in wireless environment has been an active research area. Main reason behind performance degradation of TCP is not having ability to detect actual reason of packet losses in wireless environment. In this paper, we are providing a simulation results for TCP-P (TCP-Performance). TCP-P is intelligent protocol in wireless environment which is able to distinguish actual reasons for packet losses and applies an appropriate solution to packet loss. TCP-P deals with main three issues, Congestion in network, Disconnection in network and random packet losses. TCP-P consists of Congestion avoidance algorithm and Disconnection detection algorithm with some changes in TCP header part. If congestion is occurring in network then congestion avoidance algorithm is applied. In congestion avoidance algorithm, TCP-P calculates number of sending packets and receiving acknowledgments and accordingly set a sending buffer value, so that i...

Shodhshauryam, International Scientific Refereed Research Journal © 2018 SISRRJ | Volume 1 | Issue 4 | ISSN : 2581- 6306 TCP-Performance for Wireless Environment Ranjeet V. Bidwe, Amol S. Kamble Department of Computer Engineering, PICT, Pune, MESCOE, Pune, Maharashtra, India ABSTRACT Improving the performance of the transmission control protocol (TCP) in wireless environment has been an active research area. Main reason behind performance degradation of TCP is not having ability to detect actual reason of packet losses in wireless environment. In this paper, we are providing a simulation results for TCP-P (TCP-Performance). TCP-P is intelligent protocol in wireless environment which is able to distinguish actual reasons for packet losses and applies an appropriate solution to packet loss. TCP-P deals with main three issues, Congestion in network, Disconnection in network and random packet losses. TCP-P consists of Congestion avoidance algorithm and Disconnection detection algorithm with some changes in TCP header part. If congestion is occurring in network then congestion avoidance algorithm is applied. In congestion avoidance algorithm, TCP-P calculates number of sending packets and receiving acknowledgments and accordingly set a sending buffer value, so that it can prevent system from happening congestion. In disconnection detection algorithm, TCP-P senses medium continuously to detect a happening disconnection in network. TCP-P modifies header of TCP packet so that loss packet can itself notify sender that it is lost. This paper describes the design of TCP-P, and presents results from experiments using the NS-2 network simulator. Results from simulations show that TCP-P is 4% more efficient than TCP-Tahoe, 5% more efficient than TCP-Vegas, 7% more efficient than TCP-Sack and equally efficient in performance as of TCPReno and TCP-New Reno. But we can say TCP-P is more efficient than TCP-Reno and TCP-New Reno since it is able to solve more issues of TCP in wireless environment. Keywords : TCP, Wireless networks, Protocol Design. I. INTRODUCTION are signs of congestion in the network. Flow control scheme of TCP is window based, which uses two Transmission Control Protocol (TCP) [1] is one of sender-side state variables congestion window the important standards in the internet world and (cwin) and the slow-start threshold (ssthresh). Both also a very vital element in internet protocol suite. these values are measured in bytes. A TCP sender It provides a connection oriented service with first determines values of advertised window reliable data transfer over the unreliable underlying (declared by the receiver) and the congestion protocols. TCP uses three way handshaking for window, and sends amount of data according to connection oriented service. First it establishes minimum value between them to avoid congestion. connection, then it transfers data, then it terminates connection. It uses sequence numbering and timers TCP uses a sliding window mechanism to accomplish reliable, in-order delivery, flow control to ensure reliable transfer of packets. TCP’s flow and congestion control. The basis of TCP congestion control increases the data sending rate until there control lies in the following algorithms: slow start, SHISRRJ18145 | Received : 03 Nov 2018 | Accepted : 23 Nov 2018 | November-December-2018 [ 1 (4) : 11-16] 11 S. A. Orlov et al. Int S Ref Res J November-December-2018 1(4) : 11-16 congestion avoidance, fast retransmit and fast network. Congestion can be occurred because of silly recovery. In TCP, the error recovery strategies window syndrome or less availability of bandwidth employ the congestion control scheme in the in network Second issue it is facing is Bandwidth [6]. following way. TCP continually measures how long the acknowledgments take to return. It maintains a Wireless links are having very less bandwidth as compared to the wired links (about 2 MBPS, while running average of this delay (Round Trip Time- wired links offers 10- 100 MBPS) As wireless links RTT) and an estimate of the expected deviation in offer very low bandwidth, Optimum use of available delay from the average (Delay Variation). If the bandwidth is a major issue in heterogeneous current delay is longer than the average by more networks that has to be taken care of.. Another issue than four times the expected deviation, TCP it is facing is about Mobility and handoff [7]. employs congestion avoidance as a way to deal with Nowadays, large numbers of mobile devices are lost packets.TCP works reliably well in wired added in wireless environment. When wireless host networks, where packet lost is due to congestion is moving in a particular network, its base station is only. But when it comes to wireless network, it sending data to it. But when it moves to another faces different issues like Congestion in network, station during handoff, the data sent by old base Disconnection in connections, Available Bandwidth in wireless environments, Mobility and Handoff, station is lost as it would be moved out of range. Similarly data it is sending to old base station is lost. Random packet losses, High Bit error rates, big Also the packets will experience a higher delay Round Trip Time(RTT) etc.[2], [3], [4]. during handoff due to packet rerouting. Also a random packet loss [8] is also a serious headache for TCP. TCP works reliably well on wired networks and fixed topologies, so it operates on assumption that packet lost is due to congestion. But in case of wireless environment, there are many reasons of Fig 1: TCP Window Management packet loss like disconnection, corruption by underlying physical medium, handoffs, but TCP Different proposed versions of TCP like Freeze- assumes it as due to congestion in network. So it is TCP[9], TCP-Reno [11], TCP Westwood [13], TCP- not able to find actual reason behind loss of data. Jersey [14], TCP-Westwood+ [15], and TCP-F [16], Because of this wrong assumption of TCP, it leads to TCP-Bus Algorithm [17] try to provide solution to degrade the TCP performance. Also TCP is facing mentioned issues and update changes in original TCP some another issues regarding high Bit Error Rate named as TCP-Tahoe [10]. (BER) and big round trip time (RTT). II. TCP-P APPROACH Fig 2: TCP Slow Start These are some issues currently TCP is faces: - Main issue TCP is faces due to increasing congestion [5] in network. When numbers of packets in network are more than its capacity then congestion may occur in A new designed protocol TCP-P [18] integrates working of Freeze-TCP, TCP Westwood and provides a solution for randomly lost packets. Freeze TCP works very well during handoffs or during disconnections. Freeze TCP is not able to detect or solve any other issue regarding to TCP. TCP Volume 1 | Issue 4 | November-December-2018 | www.shodhshauryam.com 12 S. A. Orlov et al. Int S Ref Res J November-December-2018 1(4) : 11-16 Westwood works well during congestions; in fact it data is delivered to the TCP receiver. The BWE value is congestion avoidance algorithm and provides a is computed by performing end-to-end estimate of faster recovery for a system. But both of the systems the bandwidth available along with the TCP are not able to detect randomly lost Packets. TCP-P is able to detect lost packets by getting notification connection by measuring and averaging the rate of returning ACKs. This estimated BWE value is used to from lost packets itself. set congestion window (cwin) and slow start TCP-P stands for TCP-Performance. It satisfies threshold (ssthresh) before congestion episode. true End-to-End semantics of TCP since no Whenever sender perceives a packet loss (i.e. a intermediaries reliable, timeout occurs or 3 duplicate ACKs are received), the connection oriented service for mobile nodes. TCP-P sender uses the BWE to properly set the congestion uses the standard TCP mechanisms for flow control window (cwin) and the slow start threshold (ssthresh) and connection management. Working of TCP-P and sends data accordingly. This mechanism is little consists of mainly three functions. First function of bit different from slow start mechanism. In slow start TCP-P deals with issues like mobility and handoff, mechanism, Disconnections. It is a receiver modifying stage, here exponentially i.e. if 2 packets delivered successfully, receiver senses wireless medium continuously for detecting fading signals which in turn detects then it will try for 4 packets, then try for 8 and so on. But in this approach packet sending rate is not happening disconnection. In certain cases, it might increased exponentially i.e. If 2 packets are delivered even be able to predict a temporary disconnection successfully and network is not able to deliver 4 (signal strength is fading for instance). In such a case, packets, and is able to send only 3 packets, then it advertises a zero window size, then it forces sender proposed approach will start sending 3 packets into the ZWP mode and prevent it from dropping its through the network, which results in preventing congestion window. When the receiver senses an system from packet loss. involved. It provides packet sending rate is increased impending disconnection, first it advertises its window size as zero and a zero window acknowledgement (called as ZWA) to sender prior to disconnection to inform sender about disconnection. This period is called as Warning Period (provided that warning period should be long enough than time required for one ZWA to get across sender). If warning period is any longer, then sender is forced into ZWP mode. If warning period is small then receiver will not have enough time to inform sender, and sender have to drop its congestion window. When connection is established again then receiver sends three ACKs for last received packet and sender starts sending data again. To check connection is Fig 4: TCP-P Approach established or not sender sends zero window probes to receiver after an interval of time. Initially TCP was designed with the notion in mind that wired networks are generally reliable and any continuously computes the connection Bandwidth segment loss in a transmission is due to congestion in the network rather than an unreliable medium (The Estimate (BWE), which is equal to the rate at which assumptions is that the packet loss caused by damage When data is being sent, at the same time, sender Volume 1 | Issue 4 | November-December-2018 | www.shodhshauryam.com 13 S. A. Orlov et al. Int S Ref Res J November-December-2018 1(4) : 11-16 is much less than 1%).This notion doesn’t hold in wireless parts of the network. Wireless links are PDR= (Total Number of Packets Delivered to Receiver)/ (Total Number of Packets Sent by Sender) highly unreliable and they lose segments all the time Here, we have calculated PDR for each version of due to a number of factors. According to [20], noise in network is main reason behind randomly lost TCP for Number of Nodes starting from 50 to 140. Following table give detailed idea about analysis. packets. Up to 30% of messages can be lost because of noise. For randomly lost packets TCP-P also provides Table 1 shows that, TCP-P is more efficient than the solution. TCP-P just modifies the header part of other TCP versions in wireless environment. the packet. When packet is lost, i.e. its lifetime Mathematically, we can say, TCP-P is 4% more exceeds TTL value of packet, and then lost packet efficient than TCP-Tahoe, 5% more efficient than itself sends loss notification message to sender. To TCP-Vegas, 7% more efficient than TCP-Sack and gain this functionality we can modify TTL field in equally efficient in performance as of TCP-Reno and TCP header to send an ICMP message to sender. This TCP-New Reno. But we can say TCP-P is more message can use value from sender IP address from efficient than TCP-Reno and TCP-New Reno since it header part of TCP to send ICMP message. By this is able to solve more issues of TCP in wireless way sender can detect lost packet and resends same packet again to receiver. environment. Another thing we can observe that, TCP-P works exceptionally well if number of nodes are equal to or greater than 100. If nodes value exceeds III.RESULTS 100, then TCP-P is most efficient protocol to serve This section is providing performance results of TCPP compared to other successful TCP versions in the system. In this case, we can say, TCP-P is more wireless environments. We have used Network Results obtained of TCP-P are also good compared to Simulator (NS-2.34) on fedora 13 system to collect other TCP versions for different network stuffs [19]. efficient than TCP-Reno and TCP-New Reno also. results. Here performance is calculated on basis of Packet Delivery ratio (PDR). PDR for a protocol can be calculated by formula: TABLE 1 : PERFORMANCE EVALUATION OF DIFFERENT TCP VERSIONS PDR (Packet delivery Ratio) for TCP. Number of Nodes 50 60 70 80 90 100 110 120 130 140 TCP-P 96.6573 94.7677 95.9261 92.8469 93.7179 96.8578 97.6224 96.9486 99.8949 96.683 TCP-New Reno 97.1467 96.3105 95.6134 92.3469 97.1445 95.5831 97.6224 96.4722 99.6662 96.683 TCP- Reno 96.0209 95.7645 95.5213 92.3459 95.5088 95.5778 96.5396 96.4722 99.8854 97.1582 Volume 1 | Issue 4 | November-December-2018 | www.shodhshauryam.com TCP-Vegas 98.7298 99.4733 95.003 93.1232 51.0204 95.2087 81.3008 96.7259 99.8788 97.7998 TCPTahoe 94.0845 95.535 95.658 88.2353 94.3602 96.4341 70.5098 94.4735 99.8271 95.0637 TCP- Sack 93.7107 93.3268 95.2294 90.6433 77.2727 90.2703 67.5676 91.8367 99.8519 97.2279 14 S. A. Orlov et al. Int S Ref Res J November-December-2018 1(4) : 11-16 Bidve College Of Engineering, Latur, India. His IV. CONCLUSION research interests lie in Computer Networking, In this paper, we can examine the issues related to wireless communication. He has presented two TCP communications over wireless environments. In frequent papers in national conferences and two papers in international conferences. He has published five disconnections, Bandwidth estimation and random papers in international journals. He is currently packet losses. Here we improve TCP performance by working as an assistant professor in Department of proposing new intelligent solution called TCP-P. 'Computer Engineering', at 'Pune Institute of particular, issues like congestion, Computer Technology, Pune'. He has total of 6 years TCP-P is an intelligent protocol which is able to detect actual reason of packet loss, i.e. congestion, of teaching and research experience. He has chaired disconnection, of random packet lost. TCP-P satisfies Committee member of various boards at reputed true end-to-end semantics as no intermediaries are Journals and conferences. many conference sessions and served as Technical not involved between sender and receiver; it just requires modification in the TCP code. TCP-P Amol Shivaji Kamble was born in controls disconnection by sensing medium continuously. It is Kolhapur, India, on 22nd March, 1987. He has completed Masters Degree in a congestion avoidance algorithm. It calculates 'Computer Engineering' from 'Pune receiving rate of ACKs from receiver and efficiently Institute rate of sending data by examining of Computer Technology, uses a bandwidth available in the medium. Also it Pune'. He has completed bachelors in Computer computes bandwidth estimate value to avoid Science & Engineering (BE CSE) in 2008 from occurring congestion. By this way, it detects 'K.I.T.'s College Of Engineering, Kolhapur, India. congestion first and tries to avoid slow start His research interests lie in Computer Networking, mechanism. It modifies header part of packet, which wireless communication. He has presented two informs sender by sending a loss notification message papers in international conferences. He is currently if packet is lost. As per our simulation results, we can working as an assistant professor in Department of use TCP-P instead of TCP-Reno or TCP-New Reno. Although this system gives similar results as TCP- 'Computer Engineering', at 'MES's College Of Reno or TCP-New Reno, but it is able to solve more teaching. ENgineering, Pune'. He has total of 6.10 years of issues of TCP in wireless environment. Also we can conclude that, if network is containing more than 100 nodes, then TCP-P is the efficient transport protocol for particular network. VI. REFERENCES [1] Forozen, Tcp/Ip Protocol Suite, 3rd ed. Tata McGraw-Hill Education,2005. V. Acknowledgment [2] Ranjeet Vasant Bidwe, "Different Issues and Survey of Proposed Solutions in TCP over Ranjeet Vasant Bidwe was born in Wireless Environments," International Journal Saswad, Pune, India, on 2nd March, of Future Computer and Communication vol. 1989. He has completed Masters Degree 2, no. 6, pp. 556-561, 2013. in 'Computer Engineering' from 'Pune [3] Institute of Computer Technology, Ranjeet V. Bidwe, Kandekar."TCP Mr. for Pune'. He has completed bachelors in Computer Environments", International Science & Engineering (BE CSE) in 2010 from 'M. S. Advanced Volume 1 | Issue 4 | November-December-2018 | www.shodhshauryam.com Engineering, Mahesh A. Wireless Journal of Management and 15 S. A. Orlov et al. Int S Ref Res J November-December-2018 1(4) : 11-16 Science(ISSN: [4] 2454-1311),vol.2,no. 11, enhanced and networking, ACM, 2001. Protocol Innovation Variants in Wireless [14] K Xu, Y Tian, TCP-Jersey for wireless IP communications, Selected Areas in Communications„ IEEE, 2004. and [15] D Kliazovich, F Granelli, TCP Westwood+ Communication (IJRITCC), ISSN: 2321-8169, enhancement in high speed long-distance PP: 5409 – 5416 networks, ICC’06.IEEE, 2006. V. Trends Jacobson, in Computing congestion control,Computer avoidance [16] K. Chandran, S. Raghunathan, S. Venkatesan, Review and R. Prakash, A feedback-based scheme for Communication dl.acm.org, 1995. improving TCP performance in ad hoc wireless Y Tian, K Xu, N Ansari, TCP in wireless networks, environments: Magazine, vol. 8, no. 1, IEEE, 2001. problems and solutions, Communications Magazine, IEEE, 2005. [8] Personal Communications [17] D Kim, CK Toh, TCP-BuS: Improving TCP Vassilis Tsaoussidis, Ibrahim Matta, Open Issues on TCP for Mobile Computing, Mobile performance in wireless ad hoc networks, Journal of Communications, mmlab.snu.ac.kr, Computing, Wiley On line Library, 2002. 2001. Ka-cheong leung and Victor O. K. LI, Transmission [9] links, Ranjeet V. Bidwe, “Different Transmission 9 , International Journal on Recent and [7] wireless international conference on Mobile computing Environments”, September 15 Volume 3 Issue [6] over pp.1905-1912,2016. Control [5] transport Control Protocol (TCP) [18] Ranjeet Bidwe, Amar Buchade TCP-P: A new in approach in wireless environment to solve wireless networks: Issues, Approaches and issues Challenges, International Conference on Computational Communications Surveys currently TCP facing, Second Tutorials, IEEE, 2006. Intelligence and Information Technology CIIT Tom goff, James Moronski, D. S. Phatak and 2012, Springer, 2012. Vipul Gupta, Freeze-TCP: A true end to end [19] Mr. Ranjeet V. Bidwe, “Improving Performance TCP enhancement mechanisms for mobile of TCP in Wireless Environment using TCP- environment, Joint P", International Journal on Communication Conference of the IEEE Computer and (ISSN: 2158-7558), vol.4, Issue no. 11, pp.48- Communications Societies, IEEE, 2000. 56, 2013 Nineteenth Annual [10] W. Stevens, TCP slow start, congestion [20] Shamimul Qamar, Kumar Manoj, Impact of avoidance,fast retransmit and fast recovery Random Loss on TCP Performance in Mobile algorithms, IETF, RFC 2001, 1997. Ad-hoc Networks (IEEE 802.11): A Simulation- [11] C Bai, J Liu, H Deng, Research on Optimization Algorithm of TCP Reno Based on Wireless Based Analysis, Arxiv preprint arXiv9:1002.2403,arxiv.org , 2010. Network, Springer, 2011. [12] Hanaa Torkey, Gamal Attiya, Ibrahim Z. Morsi, Modified Fast Recovery Algorithm for Performance Enhancement of TCP-NewReno, International Journal of Computer Applications, Volume 40 - Number 12, 2012. [13] Severio Moscola, Claudio casetti, Mario Gerla, M. Y. Sanadidi WESTWOOD: and Ren Bandwidth Wang, estimation TCP for Volume 1 | Issue 4 | November-December-2018 | www.shodhshauryam.com 16