Stochastic Modelling for Wireless Communication Networks-Multiple Access Methods.

Transcription

1 Stochastic Modelling for Wireless Communication etworks-multiple Access Methods. By Hassan KHALIL U.U.D.M. Project Report 2003: **** Examensarbete i matematisk statistik 20 poäng Handledare och examinator: Ingemar KAJ Maj 2003 Department of Mathematics Uppsala University

2

3 To my mother and in memory of my father

4

5 Contents Summary Acknowledgments Chapter : Introduction.. General etwork Concepts... Multiuser Channels..2. Duplexing Choices..3. etwork Topologies..3. Centralized Wireless etworks..3.2 Peer-to-peer Wireless etworks..4. Slotted systems.2. Wireless MAC Issues.2. Half-Duplex mode.2.2 Time Varying Channel.2.3 Burst Channel Errors.2.4 Location Dependant Carrier Sensing.2.4. Hidden terminal problem Exposed terminal problem Co-Interference Channel Capture Effect.3. Performance Metrics.3. Average Throughput.3.2 Average Packet Delay.3.3 Efficiency.3.4 Fairness of Access.3.5 Stability.3.6 Robustness Against Channel Fading.3.7 Battery Power Consumption.3.8 Ability to Serve Data, Voice and Video Chapter 2: Wireless MAC protocols 8 2. Classification 2.2 Distributed MAC Protocols 2.2. Aloha protocols Collision Avoidance Protocols Collision Avoidance with out-of-band Signaling Collision Avoidance with Control Handshaking Carrier Sense Multiple Access (CSMA) Protocols Distributed Foundation Wireless MAC (DFWMAC) Elimination Yield-on Preemptive Priority Multiple Access (EY-PMA) 2.3 Centralized MAC Protocols

6 2.3. Centralized Random Access Protocols Idle Signal Multiple Access (ISMA) Randomly Addressed Polling (RAP) Resource Auction Multiple Access (RAMA) Guaranteed Access Protocols Zhang s Proposal Disposal Token MAC Protocol (DTMP) Acampora s Proposal Hybrid Access Protocols Random Reservation Protocols (RRA) Packet Reservation Multiple Access (PRMA) Random Reservation Access-Independent Stations Algorithm (RRA-ISA) Demand Assignment Protocols Distributed Queuing Request Update Multiple Access (DQRUMA) Mobile Access Scheme based on Contention and Reservation for ATM (MASCARA) Chapter 3: Mathematical Techniques for the performance evaluation of multiple access protocols 2 3. The S-G Analysis or Poisson Analysis 3.2 The Markov Analysis 3.3 The Equilibrium Point Analysis EPA 3.3. Principles Application to Slotted ALOHA 3.4 Queueing Theory 3.5 Conclusion Chapter 4: Slotted ALOHA and Inhibit Sense Multiple Access with Capture Radio Effects 4.2 The Poisson analysis of Slotted ALOHA with capture 4.3 The Markov analysis of Slotted ALOHA with capture 4.3. Measures of dynamic performance etwork Stability 4.4 Inhibit Sense Multiple Access with Capture 4.4. System model as busy and idle periods Inhibit sense multiple access scheme Throughput of inhibit sense multiple access with capture Delay of inhibit sense multiple access 4.5 Conclusions

7 Chapter 5: Reservation-based Multiple Access Scheme for Wireless Communication etworks The parameter a 5.2 Reserved-based protocols 5.3 Reservation ALOHA (R-ALOHA) 5.4 Reserved Idle Signal Multiple Access (R-ISMA) protocol 5.4. Idle Signal Multiple Access (ISMA) scheme R-ISMA scheme S-G Analysis The Markov Analysis umerical results and conclusion Chapter 6: Packet Reservation Multiple Access (PRMA) Protocol PRMA Description 6.. Information categories 6..2 Feedback 6..3 Frame and slot 6..4 Channel access 6..5 Random information packets 6.2 PRMA system variables 6.2. Frame and slot duration Speech terminal Permission probability Speech delay and buffer size 6.3 Markov analysis for PRMA performance study 6.3. PRMA speech terminal model Markov model System stability measure Packet dropping probability Throughput Example system Conclusion 6.4 EPA for PRMA performance study 6.4. Terminal states and transitions System equilibrium Stability of PRMA Estimate of the state probability distribution System throughput Packet dropping probability 6.5 PRMA performance evaluation based on queuing theory 6.5. Introduction Front End Clipping in TASI systems Front End Clipping due to ALOHA access Combined results umerical results 6.6 Effects of capture on the stability and performance of PRMA 6.6. Capture model

8 6.6.2 PRMA stability and performance Applications 6.7 Conclusions Chapter7: Conclusions 85 Appendix I: Stability of PRMA 89 Appendix II: Packet Dropping Statistics 9 References 95

9 Summary The need for multiaccess protocols arises whenever a resource is shared by many independent contending users. Two major factors contribute to such a situation: The need to share expensive resources in order to achieve their efficient utilization and the need to provide high degree of connectivity for communication among independent users. Multiple access protocols are sometimes referred to, specially in the standards arena, as Medium-Access Control (MAC) protocols. MAC protocols occupy a sublayer within one the seven layers of the open systems interconnection (OSI) layer architecture. Both the MAC and Logical Link Control (LLC) sublayers are part of the second OSI layer, called the data link layer. The nature of the wireless channel provides new design options not available in systems using conventional wired channels, like low bandwidth, hidden and exposed stations, interference, capture, channel noise, and time varying channel. Low power requirements and half-duplex operation of the wireless systems add to the challenge. Wireless MAC protocols have been heavily researched and a large number of protocols have been proposed. Protocols have been devised for different types of architectures, different applications and different media (voice, video, and data). Analytical techniques for evaluating the performance of communication networks have evolved hand in hand with the progress in these networks. As it is usually focused on stochastic processes underlying the operations of communication networks, these techniques come from the area of applied probabilities, such as Markov processes and queueing theory. In turn, these techniques have been developed to meet the requirements of their applicability to performance evaluation. The dynamic behaviour of multiple-access systems includes the system instability. Then it is desirable to focus on three questions: ) How does system instability arise? 2) What are appropriate mathematical descriptions of the system model and how are the numerical solutions are obtained? 3) What are system parameters that influence the dynamics of the systems? Poisson processes and Markov processes are two tools used to study the system and present numerical results on throughput and delay. Another tool is called Equilibrium Point Analysis (EPA). The EPA is an alternative method to study more complex systems where multidimensional Markov processes may be required. It is a fluid-type approximation that it is applied to the steady-

10 state: it assumes that the system is always at an equilibrium point. Therefore. EPA does not necessitate calculating state transition probabilities of the Markov processes. This dissertation discusses the challenges in the design of wireless MAC protocols. A taxonomy for them is then developed in order to characterize common approaches and to provide a framwork within which these protocols can be compared and contrasted. This classification is based on architecure and mode of operation. Different analytical techniques for the performance evaluation of multiple access are then described and discussed. Finally, some protocols that deal with stability problems and performance analysis found in wireless networks are described using Poisson processes, Markov processes, EPA, and queueing theory.

11 ACKOWLEDGMETS I would like to thank my supervisor professor Ingemar Kaj for his guidance, encouragement and helpful discussions throughout the courses of this research. I would like also to thank Mr. Anders Berglund responsible for exchange student from France, for his helpful during this year at the University of Uppsala. Uppsala SWEDE, 5 may 2003 Hassan KHALIL

12 Chapter Introduction The need for multi-access protocols arises whenever a resource is accessed by a number of users. Two major factors contribute to such a situation: the need to share scarce and expensive resources in order to achieve their efficient utilization. The second reason contributing to the multi-access of a common resource by many independent entities is the need for communication among the entities to provide a high degree of connectivity. Multiple access methods are natural not only in communication systems but also in many other systems such as computer systems, storage facilities or servers of any kind.. General network concepts: A wireless network is comprised of devices communicating with each other using radio waves. These wireless devices are called nodes, stations or terminals. The names nodes, stations and terminals are used interchangeably... Multiuser channels: A multiuser channel refers to any channel that must be shared among multiple users. There are two different types of multiuser channels: the broadcast channel and the multiple access channel, which are illustrated in Figure.. A broadcast channel has one transmitter sending to many receivers, and thus the bandwidth and power of the transmitter must be divided accordingly. A multiple access channel has many transmitters sending signals to one receiver. The transmit power for each of the transmitters may vary, but the receiver bandwidth must be divided among the different users. Figure. Multiple Access and Broadcast Channels

13 ..2 Duplexing choices: In wireless communications systems, it is often desirable to allow the subscriber to send simultaneously information to the base station while receiving from the base station, and this effect is called duplexing. Duplexing may be done using frequency or time domain techniques. Frequency division duplexing (FDD) provides two distinct bands of frequencies for each user. The forward band provides traffic from the base station to the mobile, and the reverse band provides traffic from the mobile to the base station. Time division duplexing (TDD) uses time instead of frequency to provide both a forward and reverse link, each duplex channel has both a forward time slot and a reverse time slot to facilitate bi-directional communication...3 etwork topologies: There are two fundamental types of topologies used in wireless networks illustrated in Figure Centralized wireless networks: In this configuration, one station serves as the hub of the network and the user stations are located at the ends of the spokes. Any communication from one user station to another goes through the hub. The hub station controls the user stations and monitors what each station is transmitting Peer-to-peer networks: There are two variations of the peer-to-peer network topology: The distributed network, in which every user station has the functional capability to communicate directly with any other user station. In some systems, where users may be distributed over a wide area, any user terminal may be able to reach only a portion of the other users in the network, due to signal blockage or transmitter power limitations []. In this situation, user stations will have to cooperate in carrying messages across the network between widely separated stations. etworks designed to works in this way are called multi-hop networks. Figure.2 Wireless etwork topologies 2

14 ..4 Slotted systems: The wireless channel is said to be slotted if transmission attempts can take place at discrete instants in time (Figure.3). The user stations each have synchronized clocks and transmit a message only at the beginning of a new time slot. This prevents partial collisions, where one packet collides with a portion of another. Figure.3. Slotted and Unslotted Systems.2 Wireless communications characteristics: The properties of the wireless medium make the design of multiple access protocols different from and more challenging than wireline networks. The key characteristic of wireless transmission is spatial locality [2]. Only receivers within the transmission range of a station can receive its packets. The properties of wireless systems and their medium are:.2. Half-Duplex mode: In wireless systems, the received signal power varies with the transmitter-to-receiver distance r as β r where β is the power loss factor. For land mobile radio systems, a typical value of β is 4 [3]. The transmitted signal has much higher power than the received signal, which makes it impossible to detect a received signal while transmitting data. Thus in wireless systems it is very difficult to operate in full-duplex mode..2.2 Time varying channel: Radio waves arrive at a mobile receiver from different directions with different time delays. They combine via vector addition at the receiver antenna to give a resultant signal. As a result, the received signal power varies as a function of time. When the received signal strength drops below a certain threshold, the node is said to be in fade. 3

15 .2.3 Burst channel errors: As a consequence of the time varying channel and varying signal strength, errors are more likely in wireless transmissions. In wireline networks, the bit error rates are typically less than 6 0 and as a result the probability of a packet error is small. In contrast, wireless channels 3 may have bit error rates as high as 0, resulting in a much higher probability of packet errors..2.4 Location-dependent carrier sensing: Four cases of importance arise as a consequence of location-dependent carrier sensing: ) Hidden terminals, 2) Exposed terminals, 3) Co-channel interference and 4) Capture Hidden terminal problem [4]: The hidden terminal problem occurs because the wireless network, as opposed to other networks, does not guarantee high degree of connectivity. Two terminals can be within range of the intended receiver, but out-of-range of each other. Consider the case shown in Figure.4, Terminal A is transmitting to terminal B, ode C cannot hear the transmission from A. During this transmission when C senses the channel, it falsely thinks that the channel is idle. If terminal C starts a transmission, it interferes with the data reception at B. In this case node C is a hidden terminal to the terminal A. If the hidden terminals are not minimized, the throughput decreases and the delay increases. Figure-.4 The hidden terminal problem Exposed terminal problem: Exposed terminals are complementary to hidden terminals. An exposed terminal is one that is within the range of the sender but out of range of the destination [2]. In Figure.5, consider the case that terminal B is attempting to send to A, terminal C can hear the transmission from B. When it senses the channel, it thinks that the channel is busy. However, the transmission from C to D does not interfere with data reception at terminal A. In theory C can therefore 4

16 have a parallel conversation with node D. In this case C is an exposed terminal to the terminal B. If the exposed terminals are not minimized, the channel utilization decreases. Figure.5 Exposed terminal problem Co-Interference channel: The reception of a packet is subject to two criteria: the signal strength of the transmission must be above a threshold τ, and the signal to noise ratio must be above a threshold δ. A terminal is interfering if it is neither within the transmission range of the sender nor the receiver, but sufficiently close that in the presence of its transmission, the receiver cannot hear the transmitter s packets cleanly. Figure.6 shows an example. Figure.6 Interference Effect 5

17 Capture effect: The assumption that whenever two or more packets overlap at the receiver, all packets are lost, is overly pessimistic. In wireless systems the receiver might correctly receive a packet despite the fact that is time overlapping with other transmitted packets. This phenomenon is known as capture and it can happen as a result of various characteristics of radio systems. Most studies [5, 6, 7] considered power capture, namely the phenomenon whereby the strongest of several transmitted signals is correctly received at the receiver. Thus, if a single high powered packet is transmitted then it is correctly received regardless of other transmissions and hence the utilization of the channel increases..3 Performance metrics: Multiple access protocols are evaluated according to various criteria. Delay, throughput, fairness, support for multimedia, and stability are widely used metrics to compare multiple access protocols. Robustness against fading and battery power consumption are additional metrics used to compare wireless multiple access protocols. Following is a brief discussion of these metrics:.3. Average Delay: The delay is the time from the instant that a packet arrives at a source to the instant that it is successfully received. Thus the average packet delay is the ratio of the total delay of the packets in a very long interval to the number of packets in the interval. Delay characteristics are especially important for time-bounded services and applications such as voice and video..3.2 Average throughput: High bandwidth is a major objective of access schemes. Average throughput provides a measure for the percentage of capacity used in accessing the channel. More precisely, the average throughput is defined as the ratio of the number of packets that are successfully transmitted in a very long interval to the maximum number of packets that could have been transmitted with continuous transmission on the channel during the same interval of time..3.3 Efficiency: The efficiency is defined as the fraction of the time that useful information is being transmitted over the channel, where, by definition, the useful information consists only in information requests and replies. Thus, the efficiency is simply the ratio of time consumed by data slots carrying useful information to the total duration of the time..3.4 Fairness of access: A multiple access protocol is fair if it does not exhibit preference to any single node when multiple nodes are trying to access the channel. 6

18 .3.5 Stability: Although for some multiple access schemes, their delay and throughput properties might be satisfactory in the short term, they are quite poor when observed over a long interval of time. These schemes are unstable (eventually they reach a situation where the number of stations having packets ready for transmission becomes large and the throughput tends to zero)..3.6 Robustness against Channel Fading: The wireless channel is time varying and error-prone. Channel fading can make the link between two nodes unusable for short periods of time..3.7 Battery power consumption: Since the primary purpose of wireless networks is to serve mobile nodes, and since mobile nodes typically rely on battery power, efficient utilization of transmit and receive power is another important consideration for a multiple access protocols. In order to save the battery power, mobile nodes must operate in two modes: normal and sleep. There must be either a signalling procedure to wake up the destination, or some means whereby nodes can check the packet s destination address and discard those packets not matching their address..3.8 Ability to serve data, voice and video: With the increasing popularity of multimedia applications, desirable wireless networks must be able to provide some time-bounded services such as voice and video in addition to the mandatory data service. 7

19 Chapter 2 Wireless MAC Protocols 2. Classification: Wireless MAC protocols can be classified into two categories, distributed and centralized, according to the type of network architecture for which they are designed. Protocols can be further classified based on the mode of operation into random access protocols, guaranteed access protocols, and hybrid access protocols. Figure 2. shows a classification of wireless MAC protocols. In a random access protocol, nodes contend for access to the medium. When only one node makes a transmission attempt, the packet is delivered successfully. When multiple nodes make a transmission attempt, a collision results. odes resolve the collisions in an orderly manner according to rules defined by the contention resolution algorithm (CRA). In a guaranteed access protocol, nodes access the medium in an orderly manner, usually in round-robin fashion. There are two ways to implement these protocols. One is to use a master-slave configuration, where the master polls each node and the node sends data in response to the poll. These protocol are called polling protocols. The second is to operate in a distributed manner by exchanging tokens. Only the station with the token can transmit data. Each station, after transmitting data, passes the token to the next station. These protocols are called token-passing protocols. Figure2. Classification of wireless MAC protocols

20 Hybrid access protocols are based on request-grant mechanisms. Each node sends a request to the base station indicating how much time or bandwidth is required to send the data currently resident in its buffer. The request is sent using a random access protocol. The base station then allocates an upstream time slot for the actual data transmission and sends a grant to the node indicating that time slot. Depending on the intelligence at the BS, the hybrid access protocols can be further classified into Random Reservation Access (RRA) protocols and Demand Assignment (DA) protocols. In an RRA protocol, the BS has implicit rules for reserving upstream bandwidth. On the other hand, in a DA protocol the BS controls upstream data transmissions according to their QoS requirements. It collects all the requests from the nodes and uses scheduling algorithms to make bandwidth allocations. Hybrid access protocols and polling protocols by their mode of operation require a central node. Therefore they fall into the category of centralized MAC protocols. Random access protocols can operate in either architecture. Token passing protocols could be used as distributed MAC protocols but are not because of robustness considerations. Due to the time varying nature of the wireless channel, token loss would be common and token recovery is a huge overhead. As a result, all proposed distributed MAC protocols are random access protocols. 2.2 Distributed MAC protocols: 2.2. ALOHA protocols [8, 9]: The ALOHA system was first implemented at the university of Hawaii using the terrestrial radio medium. There are two versions of the ALOHA protocol: Pure ALOHA and Slotted ALOHA. In Pure ALOHA, stations transmit their packets any time they desire. Since the channel is broadcast channel, if the station hears only its own transmission, it assumes that no conflict occurred and the packet is considered a successfully transmitted packet. On the other hand, if the station hears something other than what it has transmitted, it assumes that its packet has overlapped with one or more other stations packets and must retransmit. To avoid continuously repeated conflicts, the retransmission delay is randomised across the transmitting devices, thus spreading the retry packet over time. In slotted ALOHA, the time is dividing into slots of duration equal to the transmission time of a single packet. Each user is required to synchronize the start of transmission of its packets to coincide with the slot boundary. When two packets conflict, they will overlap completely rather than partially, providing an increase in channel efficiency over pure ALOHA. The pessimistic assumption that a collision results in the loss of two packets is usually made in the analysis of an ALOHA protocol. Using this assumption, the maximum value of the throughput in a pure ALOHA is about 8% and in slotted ALOHA is about 36%. 9

21 2.2.2 Collision avoidance protocols: Collision Avoidance with Out-of-band signaling: Busy Tone Multiple Access (BTMA) [4] is an example of protocols that uses an out-of-band busy tone signal to prevent hidden terminals. When a terminal senses the transmission of another terminal, it broadcasts a busy tone signal to all others terminals. Any terminal that hears a busy tone does not initiate a transmission. Thus, all terminals in a 2R radius of transmitting terminal are inhibited from transmitting, where R is the range of the transmitting terminal. In the Receiver Initiated Busy Tone Multiple Access (RIBTMA) [0], the transmitter sends a packet preamble to the intended receiver. Once the preamble is received correctly, the receiver sets up an out-of-band busy tone and waits for the data packet. The transmitter, upon sensing the busy tone, sends the data packet to the destination. The busy tone serves two functions: to acknowledge the channel access request and to prevent transmissions from other terminals. In this protocol, only terminals within radius R of the receiving terminal are inhibited Collision Avoidance with Control Handshaking: Multiple Accesses with Collision Avoidance (MACA) [] uses three handshakes as a solution to the hidden terminal problem. A terminal that has data to send transmits a short Request To Send (RTS) packet. All terminals within one hop of the sending terminal hear the RTS and defer their transmissions. The destination responses with a Clear To Send (CTS) message. All terminals within one hop of the destination terminal hear the CTS and also defer their transmissions. On receiving the CTS the transmitting terminal assumes that the channel is acquired and initiates the data transmission. This handshaking mechanism does not completely solve the hidden terminal problem, but it does prevent it to a large extent. Enhancements to RTS-CTS control handshaking are Media Access Protocol for Wireless LA (MACAW) [2] and Floor Acquisition Multiple Access (FAMA) [3,4]. In these techniques there is a trade-off between the overhead of handshaking and the number of hidden terminals eliminated. MACAW suggested the use of the RTS-CTS-DS-DATA-ACK message exchange for a data packet transmission in the MACAW protocol. Two new control packets were added to the packet train: DS and ACK packets. When the transmitter receives the CTS packet from its intended destination, it sends out a DS (Data Sending) packet before it transmits the data packet. The DS packet notifies neighbour nodes of the fact that a RTS/CTS dialogue has been successful and a data packet will be sent. The ACK packet was implemented for immediate acknowledgment and the possibility of fast retransmission of collided data packets instead of upper layer retransmission. A new back-off algorithm, the Multiple Increase and Linear Decrease (MILD) algorithm, was also proposed in the paper [2] to address the unfairness problem in accessing the shared channel. In the MILD back-off algorithm, successful nodes decrease their back-off interval by one step and unsuccessful nodes increase their back-off interval by multiplying them by.5. 0

22 In FAMA, each ready node has to acquire the channel (the floor) before it can use the channel to transmit its data packets. FAMA uses both the carrier sensing and the RTS/CTS dialogue to ensure the acquisition of the floor and the successful transmission of the data packets. In [5], the Dual Busy Tone Multiple Access (DBTMA) is proposed; it uses both the RTS/CTS dialogue and the busy tone mechanism. Two out-of-band busy tones are used to notify neighbour nodes of the channel status. When a node is ready to transmit, it sets up its Transmit Busy Tone and sends out an RTS packet to its intended receiver. On reception of the RTS packet, the receiver sets up a Receive Busy Tone and waits for the incoming data packet. The Receive Busy Tone operates similarly to the busy tone of the RI-BTMA scheme. However, with the help of the Transmit Busy Tone, the performance is improved Carrier Sense Multiple Access (CSMA) Protocols: Distributed Foundation Wireless MAC (DFWMAC) [6,7]: This is the basic access protocol in the recently standardized IEEE 802. wireless LA standard. The DFWMAC protocol consists of four-way exchange, RTS/CTS/DATA/ACK. The handshaking is illustrated in Figure 2.2. When a terminal (sender) has data to transmit, it packs a random wait period. This wait period is decremented when the channel is idle. When this period expires, the terminal tries to acquire the channel by sending a RTS packet. The receiving terminal (destination) responds with a CTS packet indicating that it is ready to receive the data. The sender then completes the packet transmission. If this packet is received without errors, the destination terminal responds with an ACK. If an ACK is not received, the packet is assumed to be lost and retransmitted. If the RTS fails, the terminal attempts to resolve the collision by doubling the wait period. This contention resolution method is called binary exponential back-off (BEB). Figure 2.2 Four-way handshaking in DFWMAC

23 To give preference to a terminal trying to send an ACK, different waiting intervals are specified. A terminal needs to sense the channel idle for a Distributed Inter Frame Space (DIFS) interval before making an RTS attempt and Short Inter Frame Space (SIFS) interval before sending an ACK packet. Since the SIFS interval is shorter than the DIFS interval, the terminal sending an ACK attempts transmission before a station attempting to send data and hence takes priority. In addition to the physical carrier sensing, virtual carrier sensing is achieved by using time fields in the packets, which indicate to other terminals the duration of the current transmission. This time field is called the etwork Allocation Vector (AV) field. All terminals that hear the RTS or CTS message back off AV amount of time before sensing the channel again Elimination Yield-on Preemptive Priority Multiple Access (EY-PMA) [8]: EY-PMA is the channel access protocol used in the HIPERLA, HIPERLA is a highspeed (24 Mb/s) wireless LA standard for distributed networks. EY-PMA involves dividing up the channel time into cycles: Priority Resolution Phase, Contention Resolution Phase and Transmissions Phase. Contention phase is also divided into elimination phase and yield phase (figure 2.3). According to a state of channel, there are two channel access cycles such as channel free channel access cycle and synchronized channel access cycle. Channel free channel access cycle is composed of only transmission phase. When a station that creates packets detects idle channel, which is not used by anybody, it transmits directly. On the other hand, synchronized channel access cycle is composed of three phases such as prioritisation phase, contention phase and transmission phase. A station first senses the channel, if the channel is busy, it wait until the channel becomes idle and do channel contention again in the next synchronized channel access cycle. The priority resolution phase guarantees that only the highest-priority stations survive in that phase. The contention resolution phase selects a subset of the surviving stations. Finally, the surviving stations transmit during the transmission phase. Figure 2.3 Timing diagram for one channel access cycle using in EY-PMA The contention phase consists of two parts, the elimination phase and then the yield phase. A contending node desires to eliminate others in the elimination phase. After then, a 2

24 survivor of the elimination phase (i.e., a contending node which was not eliminated by other nodes) tries to yield the right of transmission to other survivors in the yield phase. In the elimination phase, a contending node transmits a burst to eliminate other contending nodes, and then listens to the channel for the duration of the elimination survival verification interval. A contention node survives the elimination phase and enters the yield phase if and only if it senses the channel idle during its elimination survival verification interval. In the yield phase, a contending node (a survivor of the elimination phase) listens for duration between 0 and the yield interval, and it survives if and only if it senses the channel idle during its yield listening. A survival of the yield phase enters the transmission phase and transmits its packets Centralized Random Access Protocols: Idle Signal Multiple Access (ISMA): ISMA [9] is a MAC scheme for centralized wireless communication networks. It enables an up stream and downstream to be transmitted on a shared channel. With the time chart of ISMA shown in Figure 2.4, when the shared channel is idle, the base station broadcasts a very short idle signal (IS) to terminals. IS will be broadcast continuously with a period 2a + b if the idle state of the channel is not changed. Here 2 a denotes the sum of round trip propagation delay and transceiver switching delay and b denotes the length of an IS. In response to an IS, a terminal having packets in the transmission buffer may transmit a packet with transmission probability p. In the case of collision, i.e., two or more terminals transmit packets after hearing the same IS, the BS broadcasts an IS again after the collision is over. In the case of successful transmission, the base station broadcasts an idle signal with acknowledgment (ISA) after receiving the full packet. For a downstream packet, the BS can transmit at any time when the shared channel is idle. Figure 2.4 Operation of ISMA [9] 3

25 It can be seen in Figure 2.4 that in ISMA when a collision occurs a complete packet is lost, resulting in poor efficiency. Reservation ISMA [20] avoids such collisions by using reservation packets, which are short packets. In R-ISMA (Figure 2.5), a node sends a reservation packet (RP) in response to an IS. If a collision occurs, a small reservation packet is lost. When the BS receives a reservation request, it sends a polling signal (PS) to that node. Only the polled node is allowed to transmit a data packet. Figure 2.5 Operation of an R-ISMA [20] Randomly Addressed Polling (RAP): As the MAC of wireless networks has to serve mobile nodes, which may move across the cell boundaries, handoff initiated by centralized scheme will make the system implementation complicated. At the same time, the dynamic nature of wireless networks makes distributed protocol hard to work reliably. RAP [2] is a centralized MAC with partial distributed functions such as initiation of handoff. The fundamental idea of RAP is that the base stations only poll those active nodes (with packet ready to transmit) under their own coverage. The protocol is illustrated in Figure 2.6, in the contention phase; all active nodes choose a code and transmit it in the same time slot. In the example, three codes (a, c, e) are chosen. The BS identifies all the codes that were transmitted. It then sends a poll for each code that was received. All nodes that picked that particular code transmit their data packet in response to the poll. If a single node has chosen the code, the transmission is successful and the BS sends an ACK. When more than one node chose the same code, a collision occurs and a ACK message is sent. When all the received codes have been polled, the BS starts another contention phase. In the example, code (a) and (e) was chosen by a single node and it resulted 4

26 in useful data transfer. Code (c) was chosen by two or more nodes, which result in collision and the complete transmission is lost. R-RAP [22] modifies the RAP protocol with a reservation mechanism to support integrated voice/data network services. GRAP [23] uses a super-frame consisting of (+) RAP frames where is the total number of codes. GRAPO [24] improves the efficiency of GRAP by dynamically changing the number of groups in a super-frame. The R-GRAP [25] protocol allows nodes to reserve a particular code in a specific frame of the super-frame. Figure 2.6 Illustration of the RAP protocol [2] Resource Auction Multiple Access (RAMA): RAMA [25, 26] is a random access protocol that achieves resource assignment using a deterministic access algorithm (Figure 2.7). Each node has a b -bit ID, and collision resolution is based on symbol-by-symbol transmission of this ID. In the contention phase, each node transmits its ID symbol-by-symbol. The BS broadcasts the symbol it heard to all the nodes. If this symbol does not match the symbol that the node transmitted, it drops out of contention. Figure 2.7 RAMA protocol Consider an example where node A with ID 0 and node B with ID 0 are contending for access to the channel. In the first round both A and B transmit and the BS acknowledges. In the second round, A transmits and B transmits 0. The channel 5

27 performs an OR operation on the signals transmitted. As a result the BS receives and acknowledges, Thus B drops out of the contention. This process continues till the complete ID is sent. An exactly b round later, the node with the highest ID always wins the contention; it then transmits the data packet Centralized guaranteed Access Protocols: The polling protocol is the only class of guaranteed access protocol that has been studied in the context of wireless networks. The main design goal of the polling protocols is to minimize the waste of bandwidth Zhang s proposal: In this protocol [27], the BS polls all the nodes in a round robin fashion for transmission requests. The protocol operation is illustrated in Figure 2.8. A node responds with a request R when it has outstanding data, or a short response S when the queue is empty, the BS then polls nodes for data according to the requests received. Figure 2.8 Zhang s proposal [27] In [27], it is proposed that all nodes should be polled once every T seconds, where T is the coherence time of the channel. Coherence time is defined as time within which the received signal strength changes by 3 db [28] Disposal Token MAC Protocol (DTMP) [29]: DTMP modifies the poll-request-poll-data cycle proposed in Zhang s proposal with just a poll data cycle. The system architecture is based upon a star topology as shown in Figure 2.9 with a single BS and multiple stations. The BS cyclically polls each station with a token; the token contains the address of the station being polled. On receipt of token, the station has authority to transmit on the channel. The token contains a flag indicating whether or not there is a data waiting to be transmitted from the BS to the station. If there is no data to be received from the BS then the station has two options. If it wishes to send data it transmits it, otherwise it does nothing. The operation of the protocol is shown in Figure

28 Figure 2.9 Disposal Token MAC Protocol [29] Acampora s Proposal [30]: As shown in Figure 2.0, the protocol operates in three phases: Polling phase, Request phase and Data phase. In the polling phase, the BS first identifies active nodes by polling them with a codeword (unique to a node). The node remains silent if it has no packets to send. An active node echoes this codeword back if it has data to transmit. The BS then broadcasts the codeword back so that every node knows the number and order of the active nodes. In the request phase, all active nodes send their requests in order to the BS. The BS polls the nodes for data during the data transmission phase. Figure 2.0 Acampora s proposal [30] Centralized hybrid Access Protocol: Hybrid protocols bridge the space between statistical access with random access protocols and deterministic access in the polling protocols by merging the best features of both types of protocol. 7

29 Random Reservation Protocols (RRA): RRA protocols try to achieve stochastic multiplexing of data on TDMA systems. Every RRA protocol has two components: random access and reservation. All nodes that have data to transmit use a random access protocol to make their first transmission. The second factor is Reservation; describes the policy enforced by the BS to reserve uplink slots for nodes that have successfully contented for the channel. In RRA protocols, the voice source model is modelled as a two state Markov model Packet Reservation Multiple Access (PRMA): PRMA [3,32] is a technique for transmitting, over short-range radio channels, a mixture of voice packets and data packets. The PRMA protocol is organized around time frames with duration matched to the periodic rate of voice packets. In each frame, time slots are dynamically reserved for packets from active voice terminals. As a consequence, the terminals with reservations share the channel in a manner closely resembling time division multiple access. The voice packet delay is constrained to meet a specific design limit. More analyses of PRMA can be found in [33, 34]. Enhancements to PRMA are PRMA++ [35], FRMA [36] and C-PRMA [37] Random Reservation Access-Independent Stations Algorithm (RRA-ISA): RRA-ISA proposes a different access policy, using Independent Stations Algorithm [38]. In [38] decisions are taken on the basis of binary channel feedback information (collision/no collision), by assuming independence in the presence of packets at the mobile stations (hence the name ISA Independent Stations Algorithm) Demand Assignment Protocols: Demand assignment protocols try to allocate bandwidth to nodes according to their QoS requirements. Wireless ATM, where end-to-end QoS is an important design goal [39], has been the motivation for these protocols. However these protocols are valid for any centralized packet radio network. It is difficult to satisfy QoS requirements of real time applications with random access protocols, because every packet has to contend for access. The time required to resolve collisions is a function of the load in the network, and bounded jitter are difficult to guarantee. Such guarantees can be typically achieved by having a central node collect the requirements and schedule transmissions for each node so as to match the requirements against available resources. Most DAMA protocols use time-slotted channels that are divided into frames, each frame is divided into an uplink and a downlink channel. These frames can be partitioned on a slot-by-slot basis (Figure 2.), in this method, each uplink and downlink consist of a single time slot. The uplink channel is divided into the request access (RA) and data transmission (TA) sub-periods. The downlink channel is divided into the acknowledgement (ACK) and the data downstream (DD) sub-periods. A user request bandwidth using the RA sub-periods. When the 8

30 BS hears a successful request (no collision) it will notify the corresponding user through the ACK sub-periods. Successful users are then assigned bandwidth, if available, in the TA subperiods. The BS transmits downstream data to mobiles uses the DD sub-periods. The RA and ACK slots are much smaller than the data slots; hence, their time intervals are called minislots. Figure 2. Radio channel classification slot-by-slot [4] Distributed Queuing Request Update Multiple Access (DQRUMA): DQRUMA [40] is designed for a fixed length packets (e.g. ATM cells) arriving at the mobile at some bursty random rates. The uplink and downlink periods are configured on a slot-byslot basis. The uplink slot comprises a single data transmission slot (TA slot) and one or more request access slot (RA slot). Figure 2.2 shows a flow chart of the DQRUMA protocol at each mobile terminal. When a mobile terminal transmits its RA, it listens to the downlink slot for its ACK. ACK indicates that the BS has received the request. Mobile users may not transmit their data until they hear b -bit access ID in the permission slot. Upon hearing the transmission permission (b-bit ID), users may transmit their data in the next uplink time slots. This is the distributed queuing aspect of the protocol, where packets are queued at mobile s buffer until the BS serves them according to scheduling policy. DQRUMA also introduces an extra bit called the piggyback (PGBK) bit in the uplink channel. Each time a mobile transmits a packet, it also includes the (PGBK) bit to indicate whether it has more packets in the buffer. This bit serves as a contention free transmission request for a mobile transmitting a packet. The BS checks this bit and updates the appropriate entry in its request table accordingly. When the bit is included, a mobile does not need a request for channel access in the following time slot. The BS knows that the mobile has more data to transmit and will assign a time slot to the mobile accordingly; this is the update portion of the protocol. Enhancement to DQRUMA is ARCMA [4]. 9

31 Figure 2.2 DQRUMA flow chart [4] Mobile Access Scheme based on Contention and Reservation for ATM (MASCARA): MASCARA [42] is the MAC protocol designed for the MAGIC WAD project. MASCARA uses variable-length frames. Each frame consists of three periods: broadcast, reserved and contention. The broadcast period is used to tell all nodes the structure of the current time frame and the scheduled uplink transmissions. The reserved period consists of a down period in which the BS transmits the downlink data and an up period when the nodes transmit packets in the order as defined by the BS at the start of the frame. The contention period is used to send new requests to the BS. 20

32 Chapter 3 Mathematical Techniques for the Performance Evaluation of Multiple Access Protocols Among the random access protocols, Slotted ALOHA is the most basic one and also it is the component of various important protocols. In addition, the performance analysis of Slotted ALOHA is rather easy compared with that of other random access protocols. In the following, we describe the mathematical techniques in the application to Slotted ALOHA. 3. The S-G Analysis or Poisson Analysis: The S-G analysis was devised to evaluate the performance of the pure ALOHA protocol in the ALOHA SYSTEM at the University of Hawaii. Since the principle of the S-G analysis is simple, the technique has been applied to many contention-based protocols in various system configurations. In this technique, we make the following assumptions: - Traffic source consists of an infinite number of users who collectively form an independent Poisson source with a mean packet generation of S packets/time slot. 2- The sum of new packet transmissions and retransmissions in the channel (called offered load) can be approximated as a Poisson process at rate of G packets/time slot. 3- Steady-state conditions (statistical equilibrium) exist. The standard unit of traffic flow is the Erlang. Here, we can define an Erlang by thinking of the channel time being segmented into slots of T seconds each; then a traffic flow of one packet per T seconds has a value of one Erlang. By the definition of the throughput S cannot exceed Erlang without collisions, and thus we expect the throughput to be bounded as 0 < S <. We note further that if the offered traffic load is very low, there will be very few retransmissions, so that we expect S G at low traffic load. At very high traffic loads, we expect a large number of collisions and consequent retransmissions, so that we will have S << G, and S will decrease toward 0. The probability that k packets are generated during a given time slot obeys a Poisson distribution with a mean of G packets per time slot; that it is, k G G e Pr( k) = (3.) k! In the S-G Analysis, we describe the throughput S as a function of the channel traffic G. let us now consider Slotted ALOHA and apply the S-G analysis.

33 Since the throughput is the fraction of time the channel carries useful information, namely, non-colliding packets, hence, under steady-state conditions, the throughput is equal to S. In slotted ALOHA, each transmitted packet will be successful only if no other packets is transmitted in the same slot. Thus, we have: ( 0) S = G Pr (3.2) where we define P(0) as the probability of no collision, which is the probability that no other packet is generated during the vulnerable interval of one packet time (Figure 3.). Using Eq. (3.), the average rate of packet arrivals in one packet slot is G, and therefore the probability that zero packets are generated in a interval that is one packet time long is and thus the throughput is G ( 0) = e Pr (3.3) S = Ge G (3.4) By maximizing S with respect to G, we obtain the maximum throughput (channel capacity). S is maximized at G = and the channel capacity is e Figure 3. Vulnerable period and slot types with Slotted ALOHA Hence, the S-G analysis is based upon the assumptions of infinite population, Poisson input and statistical equilibrium, representing approximations to the physical situation. It has been shown by other analytical techniques that the slotted ALOHA with an infinite population under the assumptions of statistical equilibrium are achievable only for some finite time period before the system goes into saturation; that is the Slotted ALOHA system with an infinite population of users is always unstable [43,44]. Thus, the S-G analysis cannot treat the stability problem. 3.2 The Markov Analysis: The Markov analysis formulates a Markovian model of the system and obtains the stationary state occupation probability distribution by calculating its state transition probabilities. The Markov analysis can study the dynamic behaviour of the system, i.e., the behaviour taking into account the system instability. Thus, the Markov analysis is a desirable technique for the performance evaluation of multiple access protocols. However, there is a great difficulty in 22

34 applying the Markov analysis to complicated protocols, which are modelled as multidimensional Markov chains with a vast amount of entries in their state transition probability matrices. Calculation of the state transition probabilities of such Markov chains is very difficult to carry out, and solving the corresponding set of simultaneous equations is infeasible. We apply the Markov analysis to the Slotted ALOHA; the analysis is based on that of [43]. We consider a case in which Slotted ALOHA is used by a group of M users each with a single buffer packet. All packets are of the same size, requiring T seconds for transmission, which is also the slot-duration. Reference [43] uses the following packet-scheduling model (referred to as the linear feedback model) (Figure 3.2). Every user is assumed to be in one of two modes: T (Thinking) or B (Backlogged). In the thinking mode, the user generates a packet in every slot with probability σ and does not generate a packet in a slot with probability σ. The packet generation is an independent process distributed geometrically with mean. In the next σ time slot, if transmission was successful the user remains in the thinking mode and the packet generation starts again. If packet transmission was unsuccessful the user moves to the backlogged mode and schedules the retransmission of the packet according to an independent geometric distribution with parameter ν. In other words, in every time-slot the user will retransmit the packet with probability ν and will refrain from doing so with probability ν. That is the retransmission delay is geometrically distributed with mean. While in the ν backlogged mode the user does not generate any new packets. Figure 3.2 Feedback model of Slotted ALOHA [45] Let the slots of the system be numbered sequentially k = 0,,... and (k) denote the number of backlogged users at the beginning of the k th slot. The random variable (k) is referred to as the state of the system. (k) is a discrete-time Markov chain. The discrete state space will consist of the set of integers 0,,2,..., M. { } The transition diagram for the system is shown in Figure 3.3. Upward transitions are possible between every state and all the higher-numbered states, since collision of any number 23