CSMA/IC: A New Class of Collision free MAC Protocols for Ad Hoc Wireless Networks Tiantong You (you@cs.queensu.ca) Department of Computing and Information Science Chi-Hsiang Yeh (yeh@ece.queensu.ca) Department of Electrical and Computer Engineering Hossam Hassanein (hossam@cs.queensu.ca) Department of Computing and Information Science Queen s University Kingston, ON K7L 3N6, Canada Abstract In this paper, we present a new collision-free MAC protocol Carrier Sense Media Access with ID Countdown (CSMA/IC) for ad hoc wireless networks that can achieve 1% collision-free performance by solving the hidden terminal problem and the concurrent sending problem. Compared to CSMA/CA of IEEE 82.11, it also improves the network's performance in decreasing the average delay of packet sending, dropping discard ratio, and increases the network's throughput significantly. Furthermore, it can enable different packets with different priority to access the media and thus gain QoS. I. INTRODUCTION In the ad-hoc [1] type of wireless network, portable devices are brought together to form a network on the fly. Usually every node is able to communicate with every other node when all nodes are spread around a relatively small geographic range. However, nodes may spread over a larger geographic range than the communication signal can reach. In this case, nodes may have to communicate over multiple hops. There is only one medium that is shared by all the nodes that are in the same radio communication range, and the radio frequency bandwidth is limited. As well, packet collisions are unavoidable due to the fact that traffic arrivals are random and there is non-zero propagation time between transmitters and receivers. Therefore, Medium Access Control (MAC) schemes are used to coordinate access to the single channel in the network. Mobile terminals have to contend for medium access by themselves. Carrier Sense Medium Access (CSMA) [2] is the main mechanism to implement medium access. Consequently, transmissions of packets from distinct mobile terminals are more prone to overlap, resulting in packet losses. Retransmissions are required and a noticeable delay appears. Recently, a lot of research has been done on local wireless MAC protocols. The IEEE 82.11 [5-7] protocol, which defines the MAC and physical layer standards for license-free industrial, scientific and medical (ISM) bands [3] allocated by the Federal Communications Commission (FCC) in the USA, is the most popular MAC for wireless LANs. It consists of Distributed Coordination Function (DCF) and Point Coordination Function (PCF). The DCF is based on the CSMA/CA mechanism and solves the hidden terminal problem [4] efficiently with the Request-to-send message (RTS) and Clear-to-send message (CTS). In this case, all the neighboring nodes that heard either of these two messages will reserve the media for data packet sending. In the scheme, the transmission radius for all the packets (including data packets, function packets, etc) is fixed. To solve the concurrent sending of neighboring nodes, the IEEE 82.11 has defined three different lengths of Inter-frame spaces (IFS) and extension time slots for use between the sending of consecutive packets. That is, when the node senses the medium is idle; it does not send the packets immediately, but first senses and confirms that the medium is idle for the different time periods. Thus the higher priority packets could be given a short time period of sensing before being sent. RTS-CTS dialogues can solve the hidden terminal problem for data packets, but cannot solve the hidden terminal problem between for the transmissions of RTS and CTS messages. The length of time to sense the medium is not unique to the node before sending the packet. So it could not avoid having the same sensing length as neighboring nodes and this may lead to collision. When the collision takes place, the extension time slot number may be enlarged, leading to a large overhead when competition for the medium occurs. In this paper, we introduce a 1% collision-free MAC protocol - Carrier Sense Media Access with ID Countdown (CSMA/IC) for ad hoc wireless networks by completely solving the hidden terminal problem and totally avoiding concurrent sending by neighboring nodes. The sensing length before packet sending is reasonable and the use of RTS-CTS packets becomes unnecessary. All of these decrease overhead. This paper is organized as follows. The next section discusses the CSMA/IC protocol. The evaluation of protocols by simulation will be discussed in Section III. In Section IV, the conclusion will be made.
II. CSMA/IC PROTOCOL 2.1. Basic Mechanisms for CSMA/IC In the CSMA/IC protocol, the transmission radius is fixed to certain unified value and the sensing radius is twice or more than that of the transmission radius. This could be achieved by simply setting the noise signal threshold lower, or we could use an extra narrow band control channel that has twice of the data transmission radius be dedicated to send the so-called buzz signal. As Fig. 1 shows, the circle 1 is the maximum transmission range of node a ; the nodes outside this range or on the edge of this range cannot receive the data clearly. We suppose that the collision radius is the same as the transmission radius. The radius of circle 2 is the sensing radius which is twice or more that of radius 1. The nodes inside the sensing radius, but outside the transmission radius, can sense (not hear) the sending signal of the sender and thus suppress sending to avoid collision. Therefore, the hidden terminal problem can be solved automatically. This approach is referred to as sensitive CSMA (S-CSMA) in [8]. However, S-CSMA also makes the exposed terminal problem even more serious by blocking a larger transmission area than CSMA. This overhead of CSMA/IC could be compensated in CSMA/CP [9] where the capacity is separated into two channels a control channel and a data channel. The control channel uses CSMA/IC while the transmission of data packets is based on scheduling through the control channel and thus solves the hidden terminal and exposed terminal problems at the same time. Note that the control packets are relatively small and this reduces the overhead of single channel S-CSMA. The system needs to be synchronized so that mobilestations (MTs) could compete for the media beginning at the same time point. External devices like a satellite Global Positioning System (GPS) could be used to make the synchronization. Our CSMA/IC only requires local synchronization between neighboring competing nodes, and the asynchrony between the nodes several hops away will not affect or have trivial effect to the function of our CSMA/IC protocol. To achieve the local synchronization, every MT will maintain a clock. The MTs will only compete the media periodically depend on their own clock. To synchronize the MTs clock locally, every MT will send a so-called Synchronizing Beacon signal (SB) periodically based on the competing period. That is, whether the MT will send the SB signal this round will be decided randomly. The surrounding MTs will adjust their clock when hear the SB signal. If the MT hears two asynchronous SB signals (from nodes hidden to each other), it will send the SB signal next round based on the first SB signal previously heard, and other MTs will adjust their clock according to the new round SB signal. Before competing, the node first senses the medium, if the medium is not idle (there is node sending data), it will quit the current competing and wait for next round. We use the binary count down mechanism to achieve media access. As Fig. 2 shown, the time period is consists of competing units, and every competing unit begins by a time slot of medium sensing, then SB signal sending period (MTs may or may not send SB signal during this period), binary competing slots, and the data-sending period. For each packet that will be sent, the node will create an equal length binary number (its value must be unique in the surrounding nodes). The higher value of the binary number means higher priority. Thus we can encrypt the packet s priority (packet type and waiting time) to the binary number. Corresponding to the binary number, the Inter Frame Space (IFS) consists of slots of time that match every bit of the unique binary number. To make our CSMA/IC robust, the time slot should be sufficiently large (e.g., larger than 2 times of maximum propagation, turn-around time plus sensing time). During competition, at the time slot that represents the bit in the binary number, the node will sense the medium. When it senses that the media is busy, it will quit the competition; otherwise the node will continue competing for the medium. At the time slot that represents the bit 1 in the binary number, the node will send a buzz signal. The nodes that finish processing the entire binary number win the media and send the packet. So, in this competition, the node with the higher value of binary number always wins the media. For the binary code design, we separate the code to two parts the first part is assigned for the packets priority, while Competing units 2 1 a b ~One competing unit~ time Fig. 1 The Transmission Range and Sensing Range Data sending period Binary competing slots SB signal sending period Medium sensing slot Fig. 2 The frame format of competing unit
the second part of code is assign for the uniqueness of competence. The priority part could be create depend on packet s type, and packet s waiting type. The higher priorities packets will be given a large value binary code. We could use the unique MAC ID as the second part of the binary code that will guarantee the unique winner when the two senders have same priority packet to send. In the competition, a lower priority function packet will lose the competition and quit in the same range. In the case of same priority packets, the lower MAC ID node will lose and quit. Any receiver must return back an ACK when they receive a data packet to confirm the reception. If the sender does not receive the ACK after sending the data packet, it will regard the data packet has suffered collision and will re-schedule to send it again. If the sender tries several times without any response, the destination will be regarded as shutting down or moving out of sender s maximum transmission radius. In our proposed CSMA/IC MAC protocol, the ACK has the highest priority, because when a node sends out the data packet, it hopes the feedback can come back immediately without delay. 2.2 Performance Analysis of CSMA/IC In the ad-hoc wireless network, the medium is shared by all the MTs. The sending signal by itself dominates the signal from neighboring nodes. So, there is no way to detect the collision when the node is sending a signal. There are two sources of data collision. One source comes from the hidden terminal. The other is the concurrently sent signal. The CSMA/IC protocol can achieve 1% collision free performance with the following two thereon. As described in last section, the sensing radius in CSMA/IC is twice or more that of the collision radius. Thereon 1: All hidden terminals of a sending node could sense the signal sending of the sending node except if there is the concurrent sending in CSMA/IC. Proof: The destination of the legal packet sending must be inside the maximum transmission radius (γ). Then, the distance between the sender and destination must be smaller than γ. The hidden terminals which are the nodes that could interfere with the data reception of the destination node must be in the range of destination s collision area. So, the distance between destination and the hidden terminal must be smaller or equal to the collision radius that is equal to transmission radius (γ). So, the distance between the sender and any hidden terminal must smaller than 2γ - in the sensing range. As Fig. 1 shows, the node b is the furthest node that may interfere with the furthest receiver of node a. But node b is in the sensing range of node a - they could sense the sending of each other if they send at the different time. Thereon 2: If the nodes could sense the signal sending to each other, with the CSMA/IC, there are no two nodes or more with different binary numbers, which could win the media at the same time. Proof: We compare two different value binary numbers that are of equal bits length (for a small value binary number, the high position bit could use to make the binary number fixed bit long) bit by bit from the highest bit position to the lowest bit position. We must be able to find the highest bit position that has a different bit value. The bit value in the binary number with the higher value must be 1, and the binary number with lower value must be. At the corresponding time slot, the node with the bit value will sense the medium and sense signal sending from the node with the bit value 1 and quit. So, two or more winners with different binary codes could not co-exist at the same time. In our proposed CSMA/IC, the binary code the node used to compete for the media must be unique, thus guaranteeing one sender in the same sensing range. To show some advantages of CSMA/IC, we compare it with the CSMA/CA mechanism defined in the IEEE 82.11. In both mechanisms, before sending a packet, there is time overhead to compete for the medium. The competing overhead in IEEE 82.11 consists of IFS plus a variable number of time slots, which is randomly created from the Collision Window (CW) a parameter reflecting the traffic situation. Let s say the length of IFS is equal to 3 time-slots, and the maximum CW is 128, then the length of competing overhead of IEEE 82.11 will vary from 3+ to 3+128. In the CSMA/IC, the length of competing overhead is a constant value (16 time-slots is enough). So, the length of competing overhead in a binary count down mechanism is acceptable. Furthermore, it makes the synchronization more efficient. Because the length of competing overhead is constant, and the length of a function packet is constant, then we can slice the time into units equal to a constant competing overhead time + time to send a function packet. If we try to use the synchronization in the CSMA/CA, we have to slice the time into the sections equal to 3+128+time to send the packet. The bandwidth wasting in competing overhead is significant even when network traffic is light (especially when the size of the sending packet is small). Using the CSMA/IC, we can get collision free MAC, but to synchronize the wireless ad-hoc networks we need external devices. There is another potential trade off that wastes the medium resource in binary count down we call it multihop competing problem. That is, in one communication range, there may be no winner; the winner may lay several hops away - that would not happen in CSMA/CA defined in the IEEE 82.11 protocol. As Fig. 3 shows, when node a, b, c, d, e, and f want to access the medium at the same time, they use the binary key as shown in the figure. 111111 1111 a b 11111 c d 111 11 e f 1 Fig. 3 Scenario of Multi-hop competing problem
When the process starts, node f first senses that the medium is busy at time slot 2 and quits, node e quits at time slot 3, and so on. Finally, only node a wins the medium. In the case that shows in Fig. 4, the node s layout is the same as the Fig. 3, but if the binary code is modified to the value as shown in the figure, the performance will be different. In the second time slot, node b is beaten by node a and quits. So, node c survives even its binary code is smaller than node b. In the fourth time slot, node d is beaten by node c and quits, then node e survives and beats node f. So, in this case, even the value of the binary code decreases from node a to node f, but finally, node a, c, and e all win the competition and send at the same time without collision. The CSMA/IC protocol has high performance in single hop ad-hoc wireless networks, such as at a conference. All the nodes pack into one room; they are all located in each other s transmission range, no matter the sensing range. However, in the multi-hops ad-hoc wireless networks, the nodes are spread in large area. In the whole sensing area (which is 4 times the transmission area), only one node at most could be allowed to send at one time. The bandwidth cost is too high, especially for the large size data packet. The exposed terminal problem still exists. To overcome the shortcut mention above is list on our future research. III. 11 11 a b 11 c 11 d 11 e f 1 Fig. 4 Another scenario SIMULATION ANALYSIS In this section, we study the performance of our CSMA/IC wireless MAC protocol by simulation. We compare performance of our CSMA/IC protocol to the single channel CSMA/CA that is defined in the IEEE 82.11 standard. The performance metrics include (1) the transmission delay, which is measured from the time when the packet is created until the time the MT receives an ACK from the destination MT of this sent packet. We calculate the Average Delay (AD) of all the successfully received packets. (2) The Discard Ratio (DR), which is the ratio of the number of discarded packets to the total number of packets sent. There are several reasons for an MT to discard a packet. For example, before sending the packet, the MT checks the packet s destination to see whether it is reachable - if the packet s destination is not in the MT s reachable neighborhood MT list, the MT discards the packet without trying to send it. This kind of discarding does not occupy any medium resources. The packet may also be discarded because sending failed 3 consecutive times. When we calculate DR, we count the packets that are sent but failed and are finally discarded relative to the total number of packets that are sent (both those that are successful and those that fail). (3) Networks Throughput, which is a measure of the number of data packets sent successfully (get ACK) per second in the whole wireless LAN. (4) The Collision Rate (CR), which is the percentage of data-type packets sent that suffer collision. A packet-level simulator was developed using the Java programming language in order to monitor, observe and measure the performance of our protocol, using different input parameters. 4.1. Experimental setting In our simulation experiments, the channel capacity is set to 2 Mbps. All the MTs are assumed to be within a 4 4-unit grid. The CSMA/CA relative parameters are set as follows: the three IFS periods, SIFS, PIFS, DIFS and the slot defer time have been set to an effective length of 2, 3, 4 and 1 octet for 2 Mbps, respectively. The size of the Collision Window (CW) is between 8 and 128. When the MT suffers repeated collisions, the CW may reach 128. When an MT is successful for 4 consecutive times, the CW will be halved till it reaches 8. The parameters in CSMA/IC are set as follows: the length of total IFS consists of 16 time slots, each time slot has the same length effect as the slot time in the CSMA/CA. In our simulations, when packet transmission is attempted 3 times without receiving an ACK, the packet will be discarded. If the MT discards 3 packets in a row for the same destination, the destination MT is removed from the MT table. Hello messages are exchanged every.2 seconds. If the MT does not hear from its neighboring MT for 2 seconds, it will remove it from its MT table. Each experiment tests the behavior of the system for a given number of nodes (8 nodes) for 6 seconds. Our simulations consist of two stages: the network initiation stage, and the testing stage. During the network initiation stage, the MTs are created one at a time. When an MT is generated, it tries to establish its neighborhood by communicating with the hello message. After all specified MTs have been created for 5 seconds, the simulation then goes into the testing stage where data is collected for an additional 6 seconds. In the simulation, every MT is in one of two mobility states: moving or pausing. When in the moving state, the MT moves towards its target location determined in the last pause period, with a specific speed (randomly generated with a mean equal to Move Speed (1 grid unit/sec)). When the MT reaches its target location, it will reach its pausing state. The length of the pausing period is also randomly generated with a mean equal to a user defined input parameter, Pause Time (4 sec). During the pause period, it will determine the next target location and its moving speed. Hence, the MT s lifetime will consist of moving periods (each with its own speed and direction), and pausing periods (of different time intervals) in turn. Another variable parameter related to the mobility model is the transmission range, which is set to 1
grid units, and the sensing radius is equal to 2.1 times of transmission range in the control channel. However, in the data channel, the sensing radius is equal to the transmission radius. The length of the function packets and management packets is set to 2 octets. The length of the data packets is set to 2 octets. The mean packet arrival rate ( ) is set to 4, 8, 12, 16, 24, 32 units/second separately for every node. The queue size is set to 1 packets. When the average packets arrival rate ( ) is set to the value higher than the maximum capacity of packets that can be sent out, the queue that is used to store the data packet temporarily (capacity is set to 1 packets in our experiment) will be filled, and after that over speed packet arrival will be blocked. 4.2. Discussion of the results In the Pictures shown underneath, the value for every point is the mean value of 12 individual testing. In all of these experiments with our proposed CSMA/IC protocol, no matter how high the data arrival rate, there are no collisions taking place. In our simulation, the size of queue used to temporarily store the incoming data packets is 1. When the arrival rate reaches a certain point, the network traffic load becomes unbalanced, thus leading to heavy collisions, and the average delay is up sharply. But when the queue is full, the next incoming data packet will be blocked outside the node. So, when the data arrival rate is higher than the network traffic load unbalance point, network performance like Average Delay, Discard Ratio, and Network Throughput are similar and may increase slowly due to the uneven character of network (some node s queues may be full, but others may not be). Fig. 5 shows the AD of above two MAC mechanisms. We found that our proposed CSMA/IC protocol has lower AD than the DCF of IEEE 82.11 when the increases. When the is not bigger than 12 packets/sec in, the AD of both experiments becomes relatively low. This is because at these packet arrival rates, the system is in balance, the transmission ability is still over the packet arrival rate and the temporary packet-storing queue is in the not-full state. Fig. 9 shows the packets blocking rate at these packet arrival rates are or near. In the low data arrival rate, the AD of CSMA/CA is slightly better than our proposed CSMA/IC, because the time slot number in CSMA/CA stays very small, limiting the sending overhead. When the is increased, the system is beyond the balance point, and most of time, the packet temporary storing queue is in the full state. The packet arrival at the full state queue will be blocked. The AD is increased slowly to two different maximum values equal to the time to send 1 packets (queue capacity) with the different MAC mechanisms. Fig. 6 and Fig. 7 show the DR and CR of the two MAC mechanisms respectively. The DR of our proposed CSMA/IC is remaining at, because of our 1% collision free status while the CSMA/CA has around 15% discard ratio when the network traffic load is unbalanced. From Fig. 8, we could find that the network throughput of our proposed CSMA/IC has increased by more than 6%. IV. CONCLUSION In this paper, we proposed a new MAC protocol CSMA/IC for wireless ad hoc networks. CSMA/IC can provide 1% collision-free media access. Our simulation results showed that CSMA/IC improves DCF of IEEE 82.11 in terms of the average delay under heavy load and maximum achievable network throughput. The CSMA/IC protocol can also differentiate network service in packet sending and thus achieve QoS. References [1] B. Leiner, D. Nielson, and F. A. Tobagi, Eds., Proceedings of IEEE GLOBECOM, Special issue on packet radio networks, vol. 75, Jan. 1987 [2] L. Kleinrock and F. A. Tobagi, Packet switching in radio channels, part I Carrier sense multiple-access modes and their throughput-delay characteristics, IEEE Trans. Commun., vol. COM-23, Dec. 1975, pp. 14-1416. [3] Federal Communications Commision, Part 15-RADIO FREQUENCY DEVICES, Subpart D-Unlicensed Personal Communications Service Devices, 47 CER.1. [4] F.A. Tobagi and L.Kleinrock, Packet Switching in Radio Channels: Part II The Hidden Terminal Problem in CSMA and Busy-Tone Solution, IEEE Trans. On Communications COM 23, December 1975, pp. 1417 1433. [5] IEEE Standard for Wireless LAN LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications, IEEE Standard 82.11, 1997 [6] The Institute of Electrical and Electronics Engineers, Inc. IEEE Std 82.11 - Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) specifications, 1999 edition. [7] Benny Bing, High-Speed Wireless ATM and LANs by Artech House publishers [8] C.-H. Yeh, H. Zhou, and H.T. Mouftah, A time space division multiple access (TSDMA) protocol for multihop wireless networks with access points, Proc. IEEE Vehicula Technology Conf., May 22. [9] You, T., C.-H. Yeh, and H. Hassanein, A new class of collision-prevention MAC protocols for ad hoc wireless networks, IEEE Int'l Conf. Communications (IEEE ICC'3), May 23
45 4 35 3 25 2 15 1 5 4 35 3 25 2 15 1 Average Delay of Successfully Sent Data Packets msec 1 2 3 4 Fig. 5 Average Delay of Successfully Sent Data 5 (%) Collision Rate of Sent Data Packets 1 2 3 4 Fig. 7 Collision Rate of Sent Data Packets 16 14 12 1 8 6 4 2 (%) Discard Ratio of Sent Data Packets 1 2 3 4 18 16 14 12 1 8 6 4 2 Fig. 6 Discard Ratio of Sent Data Packets (unit/sec) Networks Throughput of Data Packets 1 2 3 4 Fig. 8 Networks Throughput of Data Packets 6 Blocking rate of Data packets 5 4 3 (%) 2 1 1 2 3 4 Fig. 9 Blocking Rate of Data Packet