Simulation Based Analysis of the Impact of Hidden Terminal to the TCP Performance in Mobile Ad Hoc Networks Abstract The hidden terminal is classified as the sending hidden terminal and receiving hidden terminal in this paper. The quantitive analysis of hidden terminal problem and simulation-based analysis of its mechanism is discussed. Through the extensive ns simulations the detail comparison for the impact of sending hidden terminal and receiving hidden terminal to the performance of TCP is also given. The results show that the hidden terminal will produce large volume of MAC packets collisions and RET packet loss, so the TCP always goes into RTO. In the same sending rate, UDP receiving hidden terminal is more harmful (drop TCP to 0%) to the TCP performance than sending hidden terminal (drop TCP to 80%). However, TCP sending hidden terminal is more harmful to the TCP performance than receiving hidden terminal. Keywords: Mobile Ad Hoc Networks, Hidden Terminal, TCP Performance, Simulation 1. Introduction Over the past few years, mobile ad hoc networks have emerged as a promising approach for the future mobile IP applications. This scheme can operate independently from existing underlying infrastructure and allows simple and fast implementation. Such features meet requirements primarily of applications for rescue operations, law enforcement and battlefield, etc. In MANETs, stations may rely on physical carrier-sensing mechanism to determine idle channel, such as in the IEEE 80.11[1] DCF (Distributed Coordinate Function). However, DCF cannot resolve the hidden station and the exposed station problems completely.
In wireless networks, interference is location dependent. Thus, the hidden terminal problem may happen frequently. Resolving hidden terminal problem becomes one of the major design considerations of MAC protocols. IEEE 80.11 DCF is the most popular MAC protocol used in both wireless LANs and mobile ad hoc networks (MANETs). Its RTS/CTS handshake is mainly designed for such a purpose. However, it has an assumption that all hidden nodes are within the transmission range of receivers (e.g. to receive the CTS packet successfully). Some nodes, which are out of the transmission range of both the transmitter and the receiver, may still interfere with the receiver. This situation happens rarely in a wireless LAN environment since there most nodes are in the transmission range of either transmitters or receivers. However, in an ad hoc network, it becomes a serious problem due to the large distribution of mobile nodes and the multihops operation. Therefore, the hidden terminal problems continue to impact the transport layer protocol performance. The main contribution of this paper is the quantitive analysis of hidden terminal problem for mobile ad hoc networks. We also give detail researches for the impact of sending hidden terminal and receiving hidden terminal to the performance TCP, which is the interaction between MAC and transport protocol. The extensive and typical simulations show the different impact of different hidden terminal problem to the performance of TCP in the different type of traffics. The rest of this paper is organized as following. In section II, we briefly review the background of DCF mechanism and hidden terminal and exposed terminal problem. Section III, we analyze the hidden terminal problem with quantitive method. The simulation experiments and results are discussed in section IV. Some related work in the literature is given in Section V. In section VI we conclude the paper.
. Background Overview.1. 80.11 DCF The distributed coordinating function (DCF) of 80.11 specifies the use of CSMA/CA to reduce packet collisions in the network. A node with a packet to transmit picks a random backoff value T b chosen uniformly from the set {0, 1, CW 1} (CW is the contention window size), and transmits after waiting for T b idle slots. Nodes exchange request to send (RTS) and clear to send (CTS) packets to reserve the channel before transmission. Both the RTS and the CTS contain the proposed duration of data transmission, which is the duration field indicates the amount of time that the channel will be utilized to complete the successful transmission of the data. Other hosts that overhear either the RTS or the CTS are required to adjust their network allocation vector (NAV), which indicates for how long the node should defer transmissions on the channel. If a transmission is unsuccessful (by the lack of CTS or the ACK for the data sent), the CW value is doubled and the lost packet is retransmitted. The limit of retransmission times of RTS is always set to seven and DATA is four. If the transmission is successful the host resets its CW to a minimum value CW min... Hidden Terminal and Exposed Terminal Problem A typical hidden terminal situation is depicted in Fig 1. Stations A and C have a frame to transmit to station B. Station A cannot detect C s transmission because it is outside the transmission range of C. Station C is therefore the hidden node to station A. Since A and C transmission areas are not disjoint, there will be packet collisions at B. These collisions make the transmission toward B problematic. Although virtual carrier sensing which is based on a twoway handshaking RTS-CTS will help to alleviate the hidden station problem, the hidden station
problem still persist in IEEE 80.11 ad hoc networks. This is due to the fact that the power needed for interrupting a packet reception is much lower than that of delivering a packet successfully. Therefore, the interference range is always larger than the transmission range. An exposed terminal make the victim node delay the transmission because of the victim node is in the interference or transmission range. In Figure, we show a typical scenario where the exposed terminal problem occurs. Let us assume that A and C are within B s transmission range, and A is outside C s transmission range. Let us assume that C is transmitting to D, and B has a frame to be transmitted to A. According to the carrier sense mechanism, B senses a busy channel because of C s transmission. Therefore, station B will refrain from transmitting to A, although this transmission would not cause interference at D. The exposed station problem may result in a reduction of channel utilization. The exposed terminal scenario is like hidden terminal scenario only with the difference discussion perspective. In Fig.1 right graph, node C is exposed terminal to node B, but on the other hand, it is also a hidden terminal to node A. 3. Analysis of the Ratio of Hidden Terminal In this section, we analyze the hidden terminal ratio or probability in MANETs. At first three radio ranges related to a wireless radio is clarified. Transmission Range is the range within which a packet is successfully received if there is no interference from other radios. Carrier Sensing Range is the range within which a transmitter triggers carrier sense detection. In IEEE 80.11 MAC, a transmitter only starts a transmission when it senses the media free. Interference Range is the range within which stations in receive mode will be interfered by an unrelated transmitter and thus suffer a loss. Current implementation always has the same value of carrier sensing range and interference range.
In this paper, the exposed terminal is considered as a part of hidden terminal problem. Therefore, we only discuss the hidden terminal problem. We classify the hidden terminal into two types. If the hidden terminal is sending node, we call it sending hidden node. Thus, if the hidden terminal is receiver node, we call it receiving hidden terminal. Fig. depicts the hidden terminal position analysis. Given three nodes: A, B, C, their distance is denote by AB =x, AC =a, BC =b. Here x, a, b>0. If x< R t, a> R s, R t <b< R s, Then C is hidden terminal to A. C is also exposed terminal to B. R t : Transmission range R s : Sensing range or Interference range Now we derive how possible the node C will become hidden terminal. If circle Θ A is transmission range of A and circle Θ B is the transmission range of B, then C is in the circle Θ B but not in the circle Θ A for becoming a hidden terminal. Therefore, the total terminal in the Θ B Θ A Θ B, denote it as H (x). cos θ = x Rtx (1) θ 1 H ( x) = π * Rs *( π * Rs * *(* Rs *sin θ )* ) () π x Using (1) replace θ in (), the result is x 1 H ( x) = π * Rs * Rs arccos( ) + 4Rs x Rs 4 In IEEE 80.11, it always has R t =50m. Obviously understanding, if the distance of two nodes (A and B) is near, hidden terminal will be low possibility because one node sensing the
hidden terminal the other will also sensing the hidden terminal. From the mathematic analysis, if x= R t, in other works if x = 50m, H (x) has the maximized value: H ( 50) = 1.91 Rs. In IEEE 80.11, it always has R s =650m, if there are n nodes in the X*Y plane, then for given a pair of nodes A and B, the maximized possible number of its hidden terminal is: Number s max. (n - )*1.91R = X *Y When X=Y and R s =650m, maximized ratio of hidden terminal to one pair nodes in the Numbermax (n - )1.91 Rs 1.91* 650 MANETs is: Ratiomax = = n n * X *Y X. Fig.4 depicts the maximized ratio of hidden terminal to one pair node (one flow) with distance of 50m in X*X mobile plane. It shows if the mobile plane is lower than 1000m, the ratio of hidden terminal is very high (about 0.8 to 1) when the distance of two node in the flow is 50m. If we define the security threshold value of 0.5, the minimized plane of mobile is about 1300mX1300m. 4. Simulations and Discussion 4.1. Basic Simulation Information In this section we will discuss the impact of hidden terminal to the TCP performance. NS.8 [14] is used in our simulation. Simulation parameters are as follows: propagation mode is TwoRayGround, omni antenna, 50 packets for Interface Queue length and DROPTAIL for queue management. Routing protocol is AODV. Largest moving plane is 1500mX1500m. Height of antenna is 1.5m. Transmission distance is 50m, and signal interference or censoring distance is 550m. The signal transmission rate is M. Other simulation parameters are given in Tab.1.
4.. Simulation Analysis of the Hidden Terminal Scenario In the first simulation there are 4 nodes. Distance between node0 and node1 is 00m and keep static. Distance between node and node3 is kept 00m. At the beginning distance between node1 and node is 50m. From 50s to 600s node and node3 moves together to right at the same speed 1m/s. The purpose of this design is for that the simulation time is equal to the distance between node1 and node. TCP flow is sending from node0 to node1 and UDP flow is sending from node to node3 with the rate of 0.mb. TCP Newreno flow starts from 5s and ends in 610s. UDP flow starts from 50s and ends in 600s. Fig.4 gives the simulation scenario. Fig.5 depicts the dynamics of packet dropped event as the function of time. These packets are dropped at MAC layer due to the MAC packet collision (COL). The graph shows that during 350s to 550s, the MAC layer collision becomes dramatically. It shows that hidden terminal will produce many packet collisions in MAC layer. From 50s to 350s, the distance between node0 and node is lower than 550m. Node0 is in the sensing rage of node. Node0 and node can sense each other and coordinate the transmission, so in this time span the collision is none. From 350s to 550s, distance between node0 to node is larger than 550m, and the distance between node1 and node is lower than 550m. It is typical hidden terminal scenario, in which node1 out of the sensing range of node and node1 is in the interference range of node. Node and node0 cannot coordinate each other and so node go on sending packets to node3, at same time node0 go on sending packets to node1, which makes collision at node1. Our analysis of simulation trace file confirms that the collision happens at
node1. After 550s the distance between node1 and node are larger than 550m. The network becomes two repartitions and node will not interference node1, so the packet collision is zero. Fig.6 shows the impacts of hidden terminal problem to TCP performance. Fig.7 depicts retransmission retry times exceeded drop events (RET) as function of time in MAC layer. In Fig.7 there are about 4 times long RET drop events in 350s to 550s. RET event means the MAC layer retry times exceed the optional times (After sending 7 times of RTS packets there are still no CTS response back and after sending DATA packet for 4 times and no ACK back). These RET dropping events lead to the TCP timeout event. A RET event may result in the upper layer protocol TCP timeout if the retransmission packet is also lost. In Fig.6 there are 4 times of Slow- Start phase because of the RTO, in which the congestion window size is dropping to 1 packet (MSS). After 350s the hidden terminal occurs, the loss of MAC layer packets increases, then TCP always go into Slow-Stare phase due to RTO. We change the direction of UDP flow in simulation scenario I, and other parameters are kept the same. Thus UDP flow is from node3 to node, not from node to node3. It is a scenario of receiving hidden terminal. It has the same results. 4.3. Impact of UDP Hidden Terminal to TCP Performance In this set of simulations, we use the static scenario to show the TCP performance at 351s in the mobile scenario. In other words, node and node3 are static and have the distance of 351m. Distance between node0 and node1 is 00m. Distance between node and node3 is 00m. TCP flow is from 5s to 160s from node0 to node1, and UDP flow is from 30s to 60s, 90s to 10s from node to node3 with the rate of 0.1mb. Fig.8 shows the simulation scenario II.
Fig.9 show the results. The results show that the TCP performance will drop to its 80% when only 0.1mb UDP traffic comes out from the hidden terminal. Changing the direction of UDP flow, it becomes the receiving hidden terminal problem. The result shows the decreasing of the TCP goodput of TCP is more (TCP is dropped to 0%). It shows that in UDP flow the receiving hidden terminal is more harmful to TCP performance than sending hidden terminal. We also do the simulation with higher UDP rate, we find if we add the rate of UDP flow, the TCP goodput will be dropped more. 4.4. Impact of TCP Hidden Terminal to TCP Performance In this set of simulations, we change the traffics between node and node3 from UDP to TCP and other parameters are kept same as simulation scenario II. Fig.10 and Fig.11 show the results. The results show that if the TCP starts from the hidden terminal node, it will grab the channel and shut out the competing TCP flow. In receiving hidden terminal problem, the channel will share between two flows and the fairness will be better. It shows that in TCP flow the sending hidden terminal is more harmful to other TCP performance than receiving hidden terminal. 5. Related Works Some researchers propose and analyze the design of MAC Protocol (RTS/CTS) and IEEE 80.11MAC utilizes these mechanisms [,3]. The performance of RTS/CTS is analyzed in paper [4,5] using Markov model.
The impact of multi-hops to TCP is researched in paper [6,7], they give the result that TCP performance deduction with the hops increasing is due to the contention of MAC layer, thus they suggest to set the TCP congestion window limit to control the channel contention. They also [8] think the mobility of nodes is the major reason of TCP performance decreasing in ad hoc networks. Some researchers try to resolve the fairness problem in MAC layer to improve the TCP performance [9,10]. The problem of large interference range has been realized in paper [11,1] and they give analysis of the network capacity. Altman [13] gives a proposal of set delayed ACK threshold to 3 or higher to improve the TCP performance but not discuss stability problem. X. Wang emphasizes the TCP fairness issues in [14]. Gopal [15] gives some research of simultaneoussend problem. 6. Conclusions Hidden terminal is classified as the sending hidden terminal and receiving hidden terminal. In the same rate, UDP receiving hidden terminal will be more harmful to the TCP performance than UDP sending hidden terminal. TCP sending hidden terminal will shut down the other TCP flow, however in contrast the receiving hidden terminal is more TCP-friendly to the other TCP flow. Using the quantitive analysis, the ratio of hidden terminals in MANETs is also given. We recommend the mobile plane is 1300mX1300m. Future work is to research for the improvement scheme to the MAC protocol or the cross-layer scheme to avoid the hidden terminal problems. References [1] IEEE, IEEE Standard for Information Technology Telecommunications and Information Exchange between Systems Specific Requirements Part 11: Wireless LAN MAC and PHY Specifications, IEEE Std 80.11-1999, IEEE, New York, 1999 [] V. Bharghavan, A. Demers, S. Shenker, and L.Zhang, MACAW: A Media Access Protocol for Wireless LANs, ACM SIGCOMM94, pp. 1-5, 1994
[3] P. Karn, MACA- a New Channel Access Method for Packet Radio in Amateur Radio 9th Computer Networking Conference, pp. 134-140, September 1990 [4] Y. Wang and J.J. Garcia-Luna-Aceves, Collision Avoidance in Multi-Hop Ad Hoc Networks, in IEEE/ACM Intl. Symposium on Modeling, Analysis and Simulation of Computer and Telecommunication Systems (MASCOTS0), 00 [5] P. Chatzimisios, V. Vitsas, and A.C. Boucouvalas, Throughput and Delay Analysis of IEEE 80.11 Protocol, in Proc. Of 5th IEEE International Workshop on Networked Applications, pp. 168-174, 00 [6] Z. Fu, P. Zerfos, H. Luo, S. Lu, L. Zhang, and M. Gerla, The impact of multihop wireless channel on TCP throughput and loss, IEEE INFOCOM003, San Francisco, March 003 [7] Z. Fu, H. Luo, P. Zerfos, L. Zhang and M. Gerla, The impact of multihop wireless Channel on TCP Performance, IEEE Transactions on Mobile Computing, Vol.4 No., March/April 005 [8] Z. Fu, X.Meng, and S. Lu, How bad TCP can perform in mobile ad-hoc networks, IEEE Symposium on Computers and Communications, Italy, Jul. 00 [9] E. Royer, S. J. Lee and C. Perkins, The effects of MAC protocols on ad hoc network communication, IEEE WCNC, Chicago, IL, Sep. 000 [10] H. Wu, Y. Peng, K. Long, S. Cheng, and J. Ma, Performance of reliable transport protocol over IEEE 80.11 wireless LAN: analysis and enhancement, IEEE INFOCOM 00 [11] J. Li, C. Blake, D. Couto, H. Lee, and R. Morris, Capacity of ad hoc wireless networks, Proceeding of ACM MobiCom01, Jul. 001 [1] S. Xu and T. Saadawi, Does the IEEE 80.11 MAC protocol work well in multihop wireless ad hoc networks? IEEE Communications Magazine, vol. 39, no. 6, Jun. 001. [13] Eitan Altman, Tania Jimenez, Novel Delayed ACK Techniques for Improving TCP Performance in Multihop Wireless Networks Personal Wireless Communications 03, Venice Italy [14] Xin Wang and Koushik Kar, Throughput Modeling and Fairness Issues In CSMA/CA Based Ad-Hoc Networks, INFOCOM05 [15] Sumathi Gopal, Dipankar Raychaudhuri, Experimental Evaluation of the TCP Simultaneous-Send Problem in 80.11 Wireless Local Area Networks, SIGCOMM05 Workshops, August -6, 005 [16] The Network Simulator NS-, http://www.isi.edu/nsnam/ns/index.html. A B C A B C D Fig. 1. Hidden Terminal problem (left graph) and Exposed Terminal problem (right graph) Rs θ a b A x B C Fig.. Hidden Terminal location analysis
Fig. 3. The ratio of hidden terminal for the given plane size Tab. 1. Simulation Parameter Values Parameters Value Time slot 0us DIFS 50us SIFS 10us RTS length 160bits CWmin 31 CWmax 103 UDP length 51 MAC header length 144bits 50m 0 1 3 00m 00m Fig. 4. Simulation scenario I Fig. 5. MAC packets dropped due to collision as the function of time Fig. 6. TCP congestion window dynamics as the function of time Fig. 7 MAC RET drop event as the function of time
351m 0 1 3 00m 00m Fig. 8. Simulation scenario II Fig. 9. TCP goodput as the function of time in sending hidden terminal problem (left graph) and receiving hidden terminal problem (right graph) Fig. 10. TCP goodput from node0 to node1 and TCP goodput from node to node3 in sending hidden terminal problem Fig. 11. TCP goodput from node0 to node1 and TCP goodput from node to node3 in receiving hidden terminal problem