Impact of FTP Application file size and TCP Variants on MANET Protocols Performance

Transcription

1 International Journal of Moern Communication Technologies & Research (IJMCTR) Impact of FTP Application file size an TCP Variants on MANET Protocols Performance Abelmuti Ahme Abbasher Ali, Dr.Amin Babkir A/Nabi Mustafa Abstract MANET is a peer-to-peer, multi-hop Mobile wireless network where noes operate as hosts, generating traffic, an Routers, carrying control an ata. Packets are transmitte in store an forwar manner from source to estination. In this research paper we analyze the impact of file size on the performance of AODV, DSR an OLSR uner TCP traffic. We use ifferent TCP Variant namely Tahoe, New-Reno, CUBIC an Stanar TCP. We foun that the file size have impact on network performance but the impact of TCP variant is not outstaning an the selection of suitable MANET protocol is more important than the selection of TCP variant. Inex Terms AODV, DSR, FTP, OLSR, TCP I. INTRODUCTION A mobile a hoc network is an autonomous collection of mobile evices (laptops, smart phones, sensors, etc.) ; that communicate with each other over wireless links an cooperate in a istribute manner; in orer to provie the necessary network functionality in the absence of a fixe infrastructure[1]. It is a peer-to-peer network where noes operate as hosts, generating traffic, an Routers, carrying control an ata. Noes that are locate within each other transmission range, can irectly communicate, otherwise intermeiate noes will act as a router an relay ata packets to their estinations. Many routing protocols have been esigne for MANETs. Routing protocols in MANETs are classifie as Proactive, Reactive an Hybri. Proactive protocols (Table-riven) maintain fresh lists of estinations an their routes by perioically istributing routing tables throughout the network so that a source can fin a route immeiately when it nees it. Optimize Link -State Routing Protocol (OLSR) is an example of proactive routing protocols in MANET.Reactive or (On-eman) fins the route on-eman by flooing the network with route request packets. AHoc-On-Deman Vector Routing protocol (AODV) an Dynamic Source Routing (DSR) are examples of Reactive protocols. Hybri protocol combines the avantages of proactive an reactive routing. Zone Routing Protocol (ZRP) is an example of a hybri protocols. The protocols stuie in this research are AODV, DSR an OLSR. In this research paper we will stuy the Impact of application file size, on the performance of the protocols uner stuy. Five ifferent file sizes will be simulate for each protocol an, the performance parameters uner stuy will be evaluate to fin out which file(s) size is appropriate for each protocol. FTP is being use as the application where all the noes in the network run multiple FTP sessions. Data will be Abelmuti Ahme Abbasher Ali, Faculty of Engineering-Elneelain University-Khartoum-Suan Dr.Amin Babkir Abelnabi Mustafa, Faculty of Engineering-Elneelain University-Khartoum-Suan transmitte from one source (FTP server) to the estine noe. When the application files size increase the volume of traffic o too. This will cause congestion on the noes that are locate near the source; which will egrae the overall performance of the network. The routing protocol shoul be able to fin an alternative route to overcome congestion near the source (FTP Server).this research will also fin out which of the protocols uner stuy are able to avoi congestion by fining an alternative route. FTP uses TCP as the transport protocol; an after etermining the best file size for the protocols uner stuy; we will stuy the impact of TCP variants; namely New Reno, CUBIC, an Tahoe on the performance of routing protocols uner stuy; to figure out the best TCP variant for the appropriate file size. Many researchers have investigate the impact of TCP variant on MANET routing Protocol. In reference [2] a comparison of the performance of TCP-Reno an TCP-Vegas over MANET ha been performe. The researchers conclue that the performance ifference between TCP-Reno an TCP-Vegas over any selecte routing protocol is irrelevant; an the selecte routing protocol is more important than the selecte TCP variant. In reference [3] a comparison stuy of TCP-Reno, TCP-New Reno an TCP-SACK has been performe. The results of the comparison is that in a high ensity network when congestions an packet error rates are very likely, TCP SACK outperforms other variants in terms of retransmission attempts, uploa an ownloa response time. With the variations of mobility rates, TCP Reno ominates other congestion control algorithms in most of the cases. In Reference [4] the authors has carrie out a comparison of performance of TCP over MANET. The TCP variant implemente were Reno, New-Reno an SACK uner ifferent scenarios relate to mobility an network size. The stuy conclue that out of the three, the SACK variant can aapt relatively well to the changing network sizes while the Reno performs most robustly in ifferent mobility scenarios. II. TCP IN MANET TCP is an en-to-en an reliable transport protocol. It was originally esigne for wire networks. There are several factors that influence TCP performance in MANET, such as ynamic topology, share meium, high Bit Error Rate (BER) an signal faing [6].One of the important aspect of TCP is congestion control[5].tcp interpret a packet loss as an inicator of congestion. In MANET packet loss can occur as a result of a link failure or other errors, but this loss will be interprete by TCP as congestion, so TCP invokes congestion mechanisms. The invoke congestion mechanisms; will ecrease the network throughput; thus egraing MANET performance [7].Many congestion control exists for TCP. A brief iscussion of the TCP variants implemente in this research is iscusse in the following sections. 13

2 Impact of FTP Application file size an TCP Variants on MANET Protocols Performance A. TCP Tahoe In non-congestion environment the congestion winow size, is the same size as the receiver s avertise one.tcp Tahoe upon a loss of a segment ecrease the congestion winows size exponentially [5]. If loss continues, TCPTahoe limit transmission to a single atagram an ouble Time-out; so that routers gain more time to clear their queues. This mechanism is calle Multiplicative Decrease Congestion Avoiance [5].Slow start mechanism is use when congestion is over. It starts the congestion winow at the size of a single segment an increase the congestion winow by one segment each time an acknowlegement arrives. The overall approach is known as Aitive Increase Multiplicative Decrease (AIMD) [5]. B. Reno an New- Reno When TCP Reno receives 3 uplicate ACK s it interprets it as a sign of a segment loss. It reacts by re-transmit the segment without waiting for timeout. This algorithm is calle Fast Re-Transmit [8].New Reno acts as Reno when receiving three 3 uplicate ACK s. it iffers from RENO in that, Unless all ata outstaning is acknowlege,new Reno will not exit fast-recovery [8]. Thus it overcomes the problem face by Reno of reucing the CWD multiples times [9]. C. CUBIC In contrast to existing TCP congestions control stanars,which moify the congestion winows growth function linearly, CUBIC use a cubic function in orer to improve the scalability of TCP over fast an long istance networks[10].it make the winows growth to be inepenent of Roun Trip Time (RTT).CUBIC is scalable an fair to TCP flows[10]. III. OVERVIEW OF AODV, DSR AND OLSR A. AHoc-On-Deman Vector Routing protocol (AODV) AODV [11] is an On Deman routing protocol an thus only initiate a route iscovery when neee. Neighbor noes learn about each other s either by broacast or a HELLO messages. When a noe wants to sen a packet it first checks for the aress of estination in it is routing table; if aress exist; it start sening packets otherwise; it will start a route iscovery process by broacasting a route request packet (RREQ). All the noes that receive the RREQ packets check if they have any packet with the same broacast ientifier an same source IP aress if they o, they will iscar the packets to avoi uplicate packets. When the estination noe receives the RREQ; it sens a route reply packet RREP to the source noe by unicast in the reverse path. When an intermeiate noe iscovers an active links isconnection; or change of topology cause by noe movement; it sens a route error message (RERR) to the affecte noes. The source noe will re-initialize Route iscovery process; if it is still nee that route. In brief AODV uses three types of control messages RREQ, RREP an RERR to implement route iscovery an maintenance processes. B. Dynamic Source Routing (DSR) Protocol DSR [12] uses source routing in which the sener knows the complete hop-by-hop route to the estination. These routes are store in a route cache. The ata packets carry the source route in their packet heaer [13] DSR protocol compose of two mechanism Route Discovery an Route maintenance.dsr uses flooing to iscover a requeste route. Route maintenance is the mechanism by which a source noe can etect any changes in the network. When change etecte the source noe; can either attempt to use any route it happens to know about the estination noe or invoke a route iscovery mechanism to fin a new route to estination. Route iscovery an Route maintenance are on-eman operations. Implementation of DSR an source routing results in a loop-free packet routing,eliminate the nee for upating routing information in the passe-by noes an allow caching of information by noes that forwaring or over hearing packets for their own future use. C. Optimize Link State Routing Protocol (OLSR) Optimize link state routing (OLSR) [14] protocol is a moification an optimization of the pure link state routing. Instea of blin flooing the network, OLSR reuces the overhea of network floos through the use of, Multipoint Relay (MPR).MPR are the selecte routers (Noes) that can forwar broacast messages uring the flooing process. Each noe selects it is MPR set from among it is one hop symmetric Neighbors; such that it covers, in terms of raio signal, all symmetric two hop noes. In OLSR each noe maintains a route to every other noe in the network. The routes are store in routing-table at each noe. OLSR has three functions: packet forwaring, neighbor sensing, an topology iscovery [14]. Packet forwaring an neighbor sensing mechanisms provie routers with information about the neighbors an offer an optimize way to floo messages in the OLSR network using MPRs. The neighbor sensing operation allows noes to isseminate local information in the entire network. Topology iscovery is use to etermine the topology of the entire network; an to construct the routing tables IV. SIMULATION SETUP In this section we will escribe how the case uner stuy has been carrie out. A network of 50 stationary noe an FTP server has been create. We simulate each protocol using five ifferent file sizes which are range between low-loa an huge-loa. We measure the impact of file size base on the metric propose. Table (1) below shows the simulation parameters an table (2) shows the color use for each file size. Table (1) Simulation parameters Simulation Parameters Simulation area size 1000mx1000m Number of Noes 50 Simulation Time 600sec Routing MAC Protocols Routing protocol stuie AODV,DSR,OLSR MAC Protocol g Data Rate(bps) 1Mbps Buffer size(bits) Mobility Moel Mobility Type Spee Pause Time Raio Characteristic Transmitte Power(w) Packet Reception Power threshol -95 Traffic an FTP Parameters Inter request Time Exponential(720) Type of Service Best effort File Size(constant) in 1000,5000, an byte 14

3 Table (2) Color Coe use for File Sizes International Journal of Moern Communication Technologies & Research (IJMCTR) Performance Metrics The performance metrics use to measure the impact of application file size are as follows A. Average En-to-En elay The packet En-to-En elay is the amount of time it takes a packet to exit from a source until it reaches it is estination. This inclues all possible elays cause by buffering uring route iscovery latency, queuing at the interface queue, retransmission elays at the MAC, propagation an transfer times.en-to-en elay is expresse in Secons. En-to-En elay is important because some applications are elay sensitive. B. can be efine as the ratio of the total ata that reaches a receiver from the sener. It is expresse as bytes or bits per secon. can be affecte by many factors such as limite banwith, network topology changes, an unreliable communication between noes C. Downloa Response Time() Measure from the time a client application sen a request to the server to the time it receives response packet, it inclue every response packet sent from the FTP server to the ftp application client. The ownloa response time is highly influence by TCP s congestion winow size. More especially, the larger the congestion winow size, the shorter woul be the file response time. D. Data Packets The number of packet roppe at the application layer while transferring ata packets V. RESULT ANALYSIS AND DISCUSSION In this section we provie results obtaine from simulation an iscuss them for each protocol. A. OLSR Analysis The figures below show simulation results obtaine for OLSR Fig (2) Average En-To-En for OLSR Fig (3) Average in Downloa Response time Table [3] below inclue all results obtaine from OLSR simulation.we note that when file size is set to bytes Downloa response time increase ramatically an it became 5 time the value for bytes. Also Data increase 3 times the value for byte but the compare to byte file is not increasing significantly. For the same reasoning byte file compare to byte is also not significant so it better to sit the file size in OLSR case to byte or less Table3-Data Analysis of ifferent file size of OLSR File Size () A. AODV Analysis Fig (1) Average for OLSR for Different File size Fig (4) Average for AODV for Different File size 15

4 Impact of FTP Application file size an TCP Variants on MANET Protocols Performance Fig (5) Average En-To-En for AODV Fig (8) Average En-To-En for DSR Fig (6) Average in Downloa Response time in AODV Table [4] below inclue all results obtaine from AODV simulation.we note that when file size is set to Downloa response time increase ramatically an it became 3 time the value for bytes. Also Data increase 2 times the value for byte but the compare to byte file is increasing 2times. For the same reasoning byte file compare to increase 2 times an ata roppe increase 3 times. It is better to set file size to 50,000bytes Table4-Data Analysis of ifferent file size of AODV File Size () A. DSR Analysis Fig (7) Average for DSR for Different File size Fig (9) Average in Downloa Response time in DSR Table5-Data Analysis of ifferent file size of DSR File Size () , , (270sec) (270sec) (186sec) (186sec) Table (4) inclue all results obtaine from DSR simulation.we note that when file size is set to bytes an bytes the DSR protocol stoppe completely after 270 an 180 sec respectively, this show that ue to congestion in noes near the source of traffic DSR faile to fin another route to a voi the congeste ones an as a result the link will be saturate an no more ata packets will be forware which will lea to congestion collapse [5]; so it better to sit the file size in DSR case to byte or less. The poor elay an throughput performances of DSR are mainly attribute to aggressive use of caching, an lack of any mechanism to expire stale routes or etermine the freshness of routes when multiple choices are available [13] An as prove in [13] Stale caches in DSR have a harmful effect on TCP performance. This will answer the research question of which protocols can avoi congestion when noes near the source (FTP Server) are congeste. It is clear that AODV an OLSR can survive congestion near the source but DSR faile to o that for the above-mentione reasons. 16

5 A. TCP Variant Results an Analysis International Journal of Moern Communication Technologies & Research (IJMCTR) Here we use the 50,000 byte file size an we use TCP variant, namely CUBIC, NEW-RENO, TAHOE an STANDARD to compare the performance of the routing protocol uner stuy. The results shown below are obtaine for each protocol Table (6) OLSR Results TCP Throughpu Variant t New Reno CUBIC Tahoe Stanar From table (6) we note that TCP New-Reno outperforme other variants in term of throughput an minimum ata roppe although the ifference is not very significant. CUBIC comes next in terms of throughput then Tahoe an Stanar with ifferent variation in elay an time. This can be attribute to the ifferent mechanism followe by the variants in response to congestion. Table (7) AODV Results TCP Throughpu Variant t New Reno CUBIC Tahoe Stanar From table (7) we note that AODV perform almost equally for all the TCP variants in terms of an Downloa Response time with minor ifferences in throughput. Data roppe is more in Tahoe an Stanar while CUBIC has minimum ata roppe an minimum ownloa response time compare to other protocols. This performance inicate that the latency time or the time require by AODV to fin a route is less than the roun trip time require by the TCP variants. Table (8) DSR Results TCP Variant New Reno CUBIC Tahoe Stanar 10429(up to 41.5(up 4.6(up to 324ec) to 324ec) 324ec) 270(up to 324ec) From table (8) we note that DSR is performing poorly in terms of elay an throughput compare to OLSR an AODV. This is mainly attribute to aggressive use of caching, an lack of any mechanism to expire stale routes or etermine the freshness of routes when multiple choices are available VI. CONCLUSION In this research paper we have stuie the impact of file size an TCP variants on the performance of AODV, DSR an OLSR.OLSR out performe AODV an DSR in term of throughput an an this is ue to it is proactive nature. We also fin out that AODV an OLSR can avoi congestion routes by fining other routes while DSR faile. New Reno has achieve the best throughput in all the protocols.this achievement is ue to the fact that New-Reno always tries to maintain a wie congestion winow. AODV performs well uner ifferent TCP variant an this proves the flexibility of AODV. We can conclue that the selection of AHoc routing protocol is important than the selection of TCP variant as the results obtaine through simulations prove. In the future we will stuy the impact of noe mobility an network size on the performance of TCP variant an ifferent file size. REFERENCES [1] K. Subir, an G, Basavargia AHoc Mobile Wireless networks Principles, protocols an Applications Taylor & Francis Group.USA 2008 Page [2] Kim.D et el A comparison of the performance of TCP-Reno an TCP-Vegas over MANETs IEEE [3] M Nazmul Islam Khan & et el A survey of TCP Reno, New Reno an SACK over Mobile A-Hoc Network International Journal of Distribute an Parallel Systems (IJDPS) Vol.3, No.1, January 2012 [4] K. Natarajan an G. Mahaevan A Comparative Analysis an performance evaluation of TCP over MANET routing protocols Journal of Wireless Network an Microsystems Vol. 4 No. 1-2, January-December, 2015 [5] C.Douglas Internetworking with TCP/IP principles, Protocols an Architecture Person Eucation. New Jersey USA 2014 Page [6] O. Bazan, U. Qureshi, M. Jaseemuin an H.M. El-Saye, "Performance Evaluation of TCP in mobile a-hoc networks," in The Secon International Conference on Innovations in Information Technology, IIT 05, Toronto, Canaa, 2005, pp [7] K. Natarajan, Dr. G. Mahaevan A Comparative Analysis an Performance Evaluation of TCP over MANET Routing Protocols Int Jr of Avance Computer Engineering an Architecture Vol. 1, No. 1, June 2011 [8] V. Jacobson. Moifie TCP congestion avoiance algorithm, Technical report, 30 Apr to the en2en-interest Mailing List, URL: ftp://ftp.ee.lbl.gov/ /vanj.90apr30.txt [9] S. Floy, T. Henerson The New- Reno moification to TCP s fast Recovery algorithm RFC 2582, Apr 1999 [10] S. Ha, I. Rhee, an L. Xu, CUBIC: a new TCP-frienly high-spee TCP variant, SIGOPS Oper. Syst. Rev., vol. 42, pp , July [Online]. Available: [11] C. E. Perkins an E. M. Royer A hoc on-eman istance vector routing. In Proceeings of the 2n IEEE Workshop on Mobile Computing Systems an Applications, pages , Feb 1999 [12] D. B. Johnson an D. A. Maltz, Dynamic Source Routing in A-Hoc Networks," Mobile Computing, e. T. Imielinski an H.Korth, Kluwer Acaemic Publishers, 1996, pp [13] Perkins, C.E.; Royer, E.M.; Das, S.R.; Marina, M.K., Performance comparison of two on eman routing protocols for a hoc networks, Personal Communications, IEEE, vol.8, no.1, pp.16,28, Feb 2001 [14] T. Clausen, P. Jacquet, A. Laouiti, P. Muhlethaler; A. Qayyum, an L. Viennot. Optimize Link State Routing Protocol, in Proceeings of IEEE INMIC, Lahore, Pakistan, December