Cross Layer Design for QoS Support in Multimedia Applications over Wireless Networks. Jaydip Sen Innovation Lab TCS Kolkata

Transcription

1 Cross Layer Design for QoS Support in Multimedia Applications over Wireless Networks Jaydip Sen Innovation Lab TCS Kolkata

2 Agenda Introduction Different approaches to Cross Layer Design Applications of Cross Layer Design Case Studies Undesirable effect of cross-layer design Conclusions

3 Cross Layer Design- Introduction Advanced applications like VOIP, web browsing, multimedia conferencing & video streaming demand Widely varying and diverse QoS guarantees Adaptability to dynamically varying networks & traffic conditions Modest buffer requirements within network High and effective capacity utilization Low processing overhead per packet High bandwidth requirements are coupled with tight delay constraints

4 Cross Layer design Introduction (contd..) The approach of any layered network is to treat the layers as different entities and perform layer specific operations independent of other layers. TCP is a point-to-point protocol that sets up a connection between two endpoints using a handshake signal; hence it cannot be used in multicast environment. QoS was not an issue for the application using layered approach for non real-time applications. FIFO buffers were used for resource handling and sharing. TCP Application UDP IP Physical and Data Link

5 Cross Layer Design- Introduction (contd..) CLD is a way of achieving information sharing between all the layers in order to obtain highest possible adaptivity of any application. This is required to meet the challenging data rates, higher performance gains and QoS requirements for various real time and non real time applications. CLD is a framework of co-operation between multiple layers to combine the resources and create a network that is highly adaptive.

6 Cross Layer Design Introduction (contd..) CLD approach allows upper layers to adapt their strategies in a more effective way to varying link and network conditions. This helps to improve the end-to-end performance. Each layer is characterized by some key parameters that are passed to its adjacent layers to help them determine the best operation modes under the current channel, network and application conditions.

7 Cross Functional Nature of CLD Wireless Networking Architecture: Connection-oriented vs Connectionless Energy efficient analysis of MANETs Traffic theory & protocols Scaling laws of large scale networks Signal processing Increasing the spectral efficiency Reducing BER Reducing transmission energy Detection and estimation algorithms for multi-access. Information Theory Developing capacity limits Designing efficient source coding and channel coding algorithms

8 Different Mechanisms Method I Packet Headers Method II ICMP Messages Method III Local Profiles Method IV Networks Services

9 Method I Packet Headers Interlayer Signaling Pipe stores the cross layer information in the wireless extension headers of IPv6 packets. It makes use of IP data packets as in-band message carriers. Hence does not need any dedicated message protocol for this purpose. Disadvantage an IP data packet can only be processed layer by layer and this leads to inefficiency.

10 Method I Packet Headers (contd..) Application Transport Network Link Physical Packet Headers

11 Method II ICMP Messages A message can be generated at any layer and propagated to any upper layer, by using holes rather than pipes as in Method I. The messages are propagated through the layers using the Internet Control Message Protocol (ICMP). This is more flexible and efficient method. Disadvantage - ICMP encapsulated messages have to pass through network layer even if the signaling is required between link and application layer. Only upward ICMP messages are reported.

12 Method II ICMP Messages (contd..) Application Transport Network Link Physical ICMP Messages

13 Method III Local Profiles Cross layer information is abstracted from related layer and stored in separate profiles within a Mobile Host (MH). Other interested layers can select profiles to fetch desired information. This is a flexible method since profile formats can be tailored to specific layers which can access information directly. Not suitable for time-stringent tasks like real-time applications.

14 Method III Local Profile (contd..) Application Transport High-Layer Profile Network Link Physical Low-Layer Profile Mobile Host

15 Method IV Networks Services Channel and link information from physical layer and link layer are gathered, abstracted and managed by WCI Wireless Channel Information Servers. Interested applications access WCI servers for their desired parameters. Interfaces have to be defined among the MH, the WCI server and the application servers.

16 Method IV Networks Services (contd..) Application Transport Network Link Physical Application Server WCI Server MH Air Network

17 Some Applications of Cross Layer Design QoS support in the wireless systems Ad Hoc networks for real-time video streaming QoS mapping architecture for video delivery in wireless network Multimedia over wireless Multi hop wireless networks

18 Cross-Layer Design of Ad Hoc Networks for Real-Time Video Streaming E. Setton, T. Yoo, X. Zhu, A. Glodsmith, and B. Girod IEEE Wireless Communications Magazine, August 2005

19 Challenges The lack of established infrastructure, the network and channel dynamics, and the nature of the wireless medium offer an unprecedented set of challenges in supporting demanding applications over ad hoc wireless networks. Wireless channel is a broadcast medium; transmissions from different nodes interfere. The quality of links vary over time and space due to interference, multi-path fading, and shadowing. Network conditions are highly dynamic as nodes join and leave the network in an unpredictable manner. Video streaming demands high bandwidth and stringent delay constraints.

20 Challenges (contd..) When packets are lost or arrive late, the picture quality suffers as decoding errors tend to propagate to subsequent portions of the video. Due to very high bit rate requirement, a media stream may congest the network significantly. It is imperative to account for the potential impact of each user on the network statistics and guarantee that the network is not operating beyond its capacity. Unfortunately most network designs do not provide mechanisms for protocol layers to optimally adapt to underlying channel conditions and specific application requirements. Meeting end-to-end performance requirements of demanding applications is extremely challenging without interaction between protocol layers.

21 Cross Layer Design Approach Interdependencies between layers are characterized and exploited by adapting to information exchanged between layers and building appropriate amount of robustness into each layer. For example, routing protocols can avoid links experiencing deep fades, or the application layer can adapt its transmission rate based on the underlying network throughput and latency. In this work, a cross-layer framework is presented that incorporates adaptation across all layers of the protocol stack: application, transport, resource allocation, and link layer. Results have shown that the proposed mechanism provide significant performance gains with manageable complexity.

22 Salient Features of the Scheme Link capacities are dynamically reallocated based on link state information and stream-based multi-path source routing. Scheduling is performed by balancing network congestion and distortion of realtime media streams. The framework includes techniques to jointly optimize error-resilient source coding, packet scheduling, stream-based routing, link capacity assignment, and adaptive link layer techniques. The optimization is carried out dynamically by all nodes, based on link state communication, to continuously adapt to the changing wireless link conditions and traffic flows.

23 Cross Layer Design Architecture Application Layer Transport Layer Secure Coding and Packetization Throughput Congestion-Distortion Optimized Scheduling Packet Deadlines Rate Distortion Preamble Network Layer MAC Layer Traffic Flows Congestion-Optimized Routing Capacity Assignment Link Capacities Link Layer Adaptive Modulation

24 Functions of Different Layers Link Layer Adaptive techniques are used to maximize the link rates under varying channel conditions. MAC Layer Assigns time slots, codes or frequency bands to each links. Network Layer Network layer operates jointly with MAC layer to determine the set of network flows that minimizes congestion. Transport Layer Congestion-Distortion Optimized Scheduling is performed to control the transmission and retransmission of packets. Application Layer It determines the most efficient encoding rate that will suite the given requirements for the application.

25 Adaptive Link Layer Technique Adaptive modulation is an efficient technique to improve the data rate of a link by adapting link layer design variables to the variations of the channel environment. The parameters for these purpose may be: Transmitter power Target BER Symbol rate Any combinations of the above Two adaptive link layer techniques are proposed: Adapting the packet length depending on the SINR and the link layer parameters to optimize throughput. For a fixed packet length, optimizing link layer parameters such as symbol rate and constellation size for maximal throughput. With larger packet length, packet error rate (PER) increases and throughput reduces due to frequent retransmissions. Smaller packets, on the other hand have larger overhead. A tradeoff is to be made.

26 Adaptive Link Layer Technique (contd..) The optimal packet length L* which maximizes throughput is given by H = size of a packet header b = number of bits per symbol P e = probability of symbol error. It depends on modulation type and link SINR. When the channel gain is high, maximum symbol rate should be used with highest tolerable constellation size. The corresponding optimal packet length increases as the channel quality improves. When SINR is low, the error rate should be decreased by adding redundancy in the transmission: Reduce the symbol rate Use a larger spreading factor or heavier coding.

27 Adaptive Link Layer Technique (contd..) A joint optimization of the packet length and link layer parameters yields the following rule of thumb: High SINR region when the channel gain is high, the maximum symbol rate should be used with the highest tolerable constellation size. The corresponding optimal packet size increases as the channel quality improves. Low SINR region the error rate should be decreased by adding redundancy in the transmission. Two methods of doing this are: Reduce the symbol rate Use a larger spreading factor or heavier coding After all parameters are optimally chosen, a point-to-point link throughput is computed using wireless channel bandwidth, SINR and a link layer design parameter.

28 Adaptive Link Layer Technique (contd..) Throughput Performance Comparison

29 Adaptive Link Layer Technique (contd..) A constant gap is observed from Shannon capacity for different types of channels. The gap could be reduced by using various channel coding techniques. For more than two nodes, the wireless medium should be shared among multiple sender-receiver pairs through a transmission strategy that coordinates medium access. For a given arrangement of nodes an achievable data rate on each link is computed for the adaptive link layer mechanism using the following equation W is the channel bandwidth, Γ is a link layer design parameter.

30 Adaptive Link Layer Technique (contd..) The time fraction during which each link will be active under a particular transmission strategy is also determined. By considering all the different transmission strategies between the senders and receivers, one can compute an achievable capacity region that characterizes the data rates that are simultaneously achievable between each source-destination pair. Only a limited number of these strategies are relevant in that they achieve rates that cannot be obtained by time-sharing between other strategies. For an ad hoc network of limited size (say ten nodes), this region can be computed exactly.* * [S. Toumpis and A.J. Goldsmith, Capacity regions for wireless ad hoc networks, IEEE Trans, on Wireless Communications, Vol 2, No 4, July 2003]

31 Adaptive Link Layer Technique (contd..) Figure in the next slide shows an example of this achievable capacity region depicting all the rate pairs simultaneously achievable between two specific sender-receiver pairs. More advanced transmission schemes will increase the achievable capacity region. A capacity as well as flow shall now be optimally assigned within this region so that traffic requests can be accommodated with minimal congestion.

32 Adaptive Link Layer Technique (contd..) Capacity region slices of an example ad hoc network (a) single-hop routing, no spatial reuse; (b) multihop routing, no spatial reuse; (c) multihop routing with spatial reuse; (d) two-level power control added to (c); (e) successive interference cancellation added to (c)

33 Joint Allocation of Capacity and Flow Given the network capacity, the network layer assigns traffic flows to each link. The closer the flow is to the link capacity, the higher will be the latency on the link. The optimal capacity and flow assignment may be computed to minimize any quasi-convex cost function, such as the maximum link utilization over the links of the network. The above metric of network congestion is a quasi-convex function of network flow (f) and capacities of the links (c). It can be minimized by a bisection algorithm that involves solving a sequence of convex feasibility problems.

34 Joint Allocation of Capacity and Flow (contd..) Typically, the flow between a sender and a receiver is split among multiple paths. Advantage: path diversity may provide higher aggregate data rate through spatial reuse of the wireless spectrum. Multiple routes have uncorrelated loss patterns which can be exploited at the application layer. Disadvantages: Use of higher number of links creates more contention at the MAC layer, and the complexity of maintaining multiple routes is higher. To asses advantages of cross-layer design, design with oblivious layers are also considered. For oblivious layers, capacity and flow are optimized independently. In oblivious layers designs, bidirectional links are established between neighboring nodes, and the minimum transmission rate supported by these links are maximized.

35 Joint Allocation of Capacity and Flow (contd..) Given this fixed capacity assignment, the flow assignment is computed by minimizing the link utilization metric (Δ (c, f)). For this design, some links are assigned wireless capacity, even though they do not have any traffic to support resulting in a waste of network resources. The effect of multipath routing is vastly different for each of the two designs. For cross-layer design, regardless of the number of paths, high rates can always be supported as resources are only allocated to active links. As an example, performances of the cross-layer design and oblivious layer design for video streaming over three paths between two nodes in a simulated wireless network are observed. In both cases the video stream is encoded at the highest sustainable rate and transmitted over UDP. The received video quality is measured in terms of Peak Signal-to-Noise Ratio (PSNR).

36 Joint Allocation of Capacity and Flow (contd..) PSNR Figures Cross layer design supports data rate up to 1.9 MBPS. Best PSNR: 39 db Oblivious design supports data rate up to 250 KBPS. Best PSNR: 33 db

37 Scheduling and Rate Allocation Chou et al have proposed an optimal scheduling algorithm based on importance of each packet of a video stream. This type of scheduling aims at maximizing the decoded video quality at the receiver while abiding with a rate constraint. In congestion-limited situations, it is beneficial to use a congestion-distortion optimized (CoDio) scheduler that limits end-to-end delay [ Setton et al]. This metric better reflects the impact of a user s transmissions on the congestion of a network. In addition, it is inherently adaptive to time-varying network conditions. CoDiO scheduler selects the most important packets in terms of video distortion reduction and transmits them in an order that minimizes the congestion created on the network. For example, I frames are transmitted with priority, whereas B frames might be dropped, and only the most important frames are retransmitted. CoDiO scheduler avoids transmitting packets in large bursts as this increases the queuing delays.

38 Determination of Optimal Operating Rate For low-latency streaming, when the transmitted rate exceeds a certain threshold, self-congestion causes too much delay in the network to meet the tight delay constraint, and the received video quality eventually degrades. With independent layers, the optimal operating rate is determined by algorithms such as TCP-friendly rate control based on end-to-end statistics. In the proposed cross-layer optimization scheme, the application layer determines the optimal rate for the video stream, based on rate-distortion characteristics, delay constraints, and current network conditions. Setton et al have derived a model that captures the impact of both encoder quantization and packet losses due to congestion on overall video quality. This model is used to determine the highest sustainable rate in conjunction with the optimization described before.

39 Determination of Optimal Operating Rate (contd..) Figure in the next slide shows decoded video quality for a sequence encoded at different rates and transmitted over six paths, according to the flow assignment performed earlier. When the encoding rate approaches the maximum aggregate data rate sustainable by the network (approximately 360 KBPS), the video quality drops as the end-to-end delay increases due to congestion. When the delay tolerance is small, this performance degradation occurs at rates well below network capacity, as even a slight increase in queuing delay affects the loss rate. In other words, delay constraints reduce the effective network capacity.

40 Determination of Optimal Operating Rate (contd..) Rate-PSNR performance for live video streaming using sixpath routing for different playout deadlines

41 A Cross-Layer Quality-of-Service Mapping Architecture for Video Delivery in Wireless Networks W. Kumwilaisak, Y. Thomas Hou, Qian Zhang, Wenwu Zhu, C.-C. Jay Kuo, and Ya-Qin Zhang IEEE Journal on Selected Areas in Communications, Vol 21, No 10 September 2003

42 Video Delivery in Wireless Networks An important issue in providing multiple QoS guarantees to video applications in wireless systems is dynamic QoS management with mobility support. A dynamic QoS management system allows video applications and the underlying prioritized transmission system to interact with each other in order to minimize service degradation in a resource constrained and time-varying wireless environment. A proper coordination mechanism should be in place between priority transmission system and video applications, since in Transmission Layer the QoS is expressed in terms of probability of buffer overflow and the probability of delay constraint violation at the link layer. Whereas, in Video Application Layer QoS is measured by mean square error (MSE) and Peak Signal-to-Noise Ratio (PSNR).

43 System Components Source video encoding module (at sender) Cross layer QoS mapping and adaptation module Link layer packet transmission module Wireless channel Adaptive wireless channel modeling module Video decoder module (at receiver)

44 Architecture of a Cross Layer QoS Management for Video Delivery over Wireless Video QoS Requirement Trans. QoS Provisioning Trans. Module Video Input Video Encoder QoS Mapping & QoS Adaptation QoS 1 QoS 2 QoS k Transmission Control Video Module Adaptive Channel Modeling Channel Feedback Video Output Video Decoder Time-Varying Non-Stationary Wireless Channel

45 QoS Mapping Technique The QoS Mapping and the Adaptation Module is the key component in this cross layer design. It is designed to optimally match the application layer QoS and the link (transmission layer) QoS. The video application layer QoS and link-layer QoS interact with each other and adapt themselves according to the wireless channel condition.

46 GOP Structure of MPEG-4 Video File Video Frame 1 Video Frame 2 Video Frame 3 Video Frame N I P P P Base Layer 1 st Enh Layer. 2 nd Enh Layer Mth Enh Layer T P TP + 1/F T P + 2/F T P + (N-1)/F Playback Deadline

47 QoS Mapping Issues Packets from the same GOP (Group of Picture) structure are to placed on the same QoS class. Loss of any video portion affects the end-to-end video quality due to interdependency within the encoding structure. Each video layer is packetized into several fixed-size packets before transmission. If the video playback frame rate at the end user is F frames/sec, and if the mobile terminal starts to play back the first video frame of GOP at time Tp, then for uninterrupted playback the video frame n in the same GOP should be received before Td(n) = Tp + ((n-1)/f)

48 QoS Mapping Issues Problem statement: Given the set of rate constraints under the priority transmission system and the expected channel service rate (which can be considered stationary in a time period corresponding to one GOP), what is the optimal mapping policy for one GOP with N scalable frames (coded in M video layers) to K priority classes such that the overall distortion is minimized.? Constraints: Source rate of video bitstreams under each priority class must not exceed the rate constraint of the corresponding priority class. Summation of rate constraints of all priority classes has to be bounded by the expected channel service rate.

49 Video Adaptation Algorithm Video coding module sets up a QoS bound r(t) in terms of expected video distortion (or expected PSNR). It then sends a transmit request (Txreq) to transmission module to set up the transmission process. Transmission module offers a set of statistical QoS guarantees for each priority class. The QoS parameters of the transmission module that provide the lowest distortion and satisfy the range of video quality requirement r(t) are chosen as the QoS parameters for the transmission. If there are no QoS parameters satisfying all the constraints, the available transmission capabilities cannot met the video quality requirement. The transmission module requests the video coding module to adjust the video quality requirement. The video coding module complies with this request and adjusts the QoS bounds and repeats the process in above step.

50 Video Adaptation Algorithm Video coding module sends the selected QoS parameters to the transmission module to set up QoS parameters of each priority. Transmission module send the acknowledgment to the video coding module after its QoS parameters are set up. The prioritized video bitstream is uploaded in the priority network based on agreed upon QoS parameters and the video layer mapping policy. If any change is detected in the service rate of the transmission channel, adaptation of the QoS parameters for both the video application and the priority network is called for. (Step 2 is activated).

51 Experimental Results The 100 video frames of CIF foreman sequence are used for simulation. Video sequence is encoded by PFGS video codec with frame rate of 10 frames/s and there are 10 frames in each GOP. The non-stationary behavior of wireless channel is simulated by randomly changing the normalized Doppler frequency and average power. The Doppler frequency is chosen from the set of {10-3, 5*10-3, 10-2 }. The average SNR of the received signal varies from 10 to 20 db. Two priority classes with strict priority scheduling is considered. The first class has the higher priority. The packet size is 200 bytes. The expected service rate of the channel is set to 380 kbps at normalized Doppler frequency of The rate constraint of the first priority class computed from the effective bandwidth and effective capacity and those obtained from simulations are close to each other over a wide range of packet loss probabilities.

52 Experimental Results (contd..) Rate constraint of a high priority class (of two priority classes) computed from wireless channel model with normalized Doppler frequency 10-2 and average power 16dB. Buffer sizes are 250 and 500 packets.

53 Experimental Results (contd..) The effect of buffer size on rate constraint is also studied. It is observed larger the buffers size, the more data rate we can transmit under the same packet loss probability guarantee. The rate constraints for priority class 2 over a wide range of packet loss probability is also studied. The rate constraint of this class with a buffer size of 250 packets is dependent on how much priority class 1 occupies the wireless link. The rate constraint of priority class 2 with a lower QoS guarantee of priority class 1 (with guaranteed packet loss probability = 10-2 and rate constraint = 109 kbps) can provide a higher transmission rate than that with a higher QoS guarantee of priority class 1 (with guaranteed packet loss probability = 10-1 and rate constraint = 54.5 KBPS).

54 Experimental Results (contd..) Rate constraint of a lower priority class (of two priority classes) and a buffer size of 250 packets based on absolute priority scheduling when the packet loss rate guarantee for class 1 is equal to 10-2 and 10-4

55 A Cross-Layer Design for QoS Support in the 3GPP2 Wireless Systems M. Bourouha, S. Ci, G. Ben Brahim, and M. Guizani IEEE Globecom 2004 Workshop

56 Cross Layer Design in 3GPP2 Wireless Systems In this work, a suite of cross-layer design modules fro QoS support in the 1xEV- DV system is proposed. The proposed design consists of several modules. The modules are: Priority Admission Control modules which admits users according to their QoS priority profiles Resource Allocation Control module which is responsible for allocating the optimal combination of all system parameters. Resource Scheduling Control module which aims at achieving a better overall throughput gain and guarantees the QoS requested by different users service levels. The mechanism is based on Effective Capacity concept.

57 QoS Provisioning Architecture Upper User QoS Profiles BS QoS Validation Control Layer Resource Scheduling Control To Users MAC Resource Allocation Control PHY Priority Admission Call Rejected Transmission Requests from Different Users

58 QoS Provisioning Architecture (contd..) Priority Admission Control: Its goal is to give more priority, during resource allocation, to users with better channel conditions. It classifies and prioritizes users transmission requests based on three major parameters: The user channel condition (CQI): users with high CQI are given high priority The user QoS profile, such as data rate, delay, loss rate etc. The amount of requested resources, which will be based on the class of service.

59 QoS Provisioning Architecture (contd..) Priority Admission Control: Scenario 1: A user with poor QoS profile or without QoS profile is requesting packet data services that are not sensitive to strict QoS requirements, the QoS mode will be in a Non-Assured mode. Such request could be either granted best effort resources according to the current CQI and the user priority or denied service. If, the user is requesting packet data services that are sensitive to minimum data rate, delay, or error rate, the request is immediately denied. Scenario 2: A user with good QoS profile (high data rate, low allowable delay etc.) requests services that are not sensitive to strict QoS requirements, the best transmission resources will be allocated. In this case, no further processing by upper modules will be required. If the user request for services that are sensitive to strict QoS requirements, the appropriate transmission resources will be allocated and the QoS service level will be in Assured QoS mode. Higher modules will be involved.

60 QoS Provisioning Architecture (contd..) Resource Allocation Control: It makes the best selection among all combinations of the overall system parameters. Some of the parameters are: Number of Walsh codes Number of time slots Modulation scheme Channel code rate The resource allocation control module allocates the optimum transmission resources to different users resource requests using their classification parameters which are: Diversity priority Type of traffic QoS profile QoS mode

61 QoS Provisioning Architecture (contd..) Resource Allocation Control: It works in two steps. Step 1: Based on user s diversity value describing the channel conditions, a lower bound of the Modulation and Coding Scheme (MCS) level is allocated to each user. Users with high diversity value ( good channel conditions) are typically assigned higher MCS level (higher order of modulation and higher data rate). Step 2: When the lower bound of the MCS level for each user has been selected based on the CQI values, the resource allocation control checks the QoS mode of each user. If the QoS mode is in the Non-Assured mode, then based on the diversity value and the available resources, the lower bound of the MCS level will be assigned and no further processing is needed. This reduces processing overhead at the BS. If the QoS mode of a user is the Assured QoS mode, then it sends the selected lower bound of the MCS level to upper modules for further QoS processing.

62 Dynamic Resource Allocation Algorithm The resource allocation algorithm uses Effective Capacity theory. The idea behind EC theory is to incorporate the concept of QoS in wireless links which was missing in Rayleigh and Rician models. By introducing two EC function- the probability of non-empty buffer and the QoS exponent of the connection, the EC link layer modeling and estimation can easily associate with QoS specification. Figure in the next slide shows the throughput using the proposed design versus a system using a fixed set of system parameters. The upper graph shows the throughput using the proposed scheme. The overall throughput using the dynamic resource allocation is improved as can be shown from the graph. The data rate is fixed at 480 kbps in the second part (fixed resource allocation).

63 Performance of Resource Allocation Algorithm System Throughput

64 A Cautionary Approach on Cross-Layer Design Vikas Kawadia and P.R. Kumar IEEE Wireless Communications Magazine February 2005

65 Undesirable Interactions in Cross Layer Design Whenever Cross Layer Design is implemented, problems like time asynchronization, instability and undesirable interactions between the layers have to be taken care of. This issue is illustrated by means of an example - Adaptive Rate MAC Layer.

66 Adaptive MAC and Minimum Hop Routing The rate adaptive protocol is a simple variation of IEEE MAC Protocol. The idea behind rate-adaptive MAC protocol is to send data at higher rates when the channel quality is good. Such higher rates are achieved by changing the modulation scheme. Such a scheme can lead to undesirable effects on the higher layers. For example, when combined with minimum-hop routing it can lead to a performance degradation.

67 Adaptive MAC & Minimum Hop Routing (contd..) The reason for the adverse impact is that minimum-hop routing chooses longer hops, for which the signal strength is lower, and thus the data rate achieved through channel quality adaptation is low. CLD can thus lead to unintended adverse interactions between the MAC and the network layer in this case. A routing protocol using some other metric is necessary when adaptive-rate MAC is used.

68 Adaptive MAC Protocol It is a modification of IEEE MAC protocol. A set of rates is available (i.e. modulation schemes), and the transmission rate can be chosen before each packet is sent. The RTS/CTS and broadcast packets are always transmitted at base rate (lowest data rate). Receiver measures the received signal strength of the RTS packet, and determines the maximum rate at which data can be received. This rate is then communicated to the sender in the CTS packet. Sender sends the subsequent ACK and DATA packets at this rate.

69 Adaptive MAC Protocol (contd..) Typically transmission to nodes which are near would occur at higher rates since the path loss attenuation of the signal is small. Lower data rates would be used for nodes located farther. If at higher layer, a minimum hop routing (say DSDV) is used, this scheme will have a very adverse effect. DSDV builds routing tables by sending hello packets that contain cumulative routing information (i.e., information that has been gathered from all the neighbors of a node). Hello packets are broadcast packets sent at the base rate and large range. Minimum hop routing thus chooses the longest possible hops on the path, which causes low received signal strength, which in turn implies a low data rate.

70 Rate Adaptive MAC Protocol (contd..) In fact, if we turn off the adaptive-rate MAC and use plain IEEE at the highest data rate (i.e., do not send any data when the channel is not good enough), we can get much better end-to-end throughput. In this case, longer hops simply do not exist, and thus minimum hop routing is forced to use a larger number of short hops, which provide higher data rates.

71 DSDV Routing Protocol DSDV chooses a small number of long hops : a lower data rate when an adaptive rate MAC is used

72 Plain IEEE Protocol Plain IEEE causes short hops of higher data rate to be used

73 Adaptive MAC Protocol Scheme 1 A fixed transmit power level of 0.28 W A receiver-transmitter distance of 0 99 m yields a data rate of 11 mbps, m yields a data rate of 5.5 mbps, a distance of m yields only 2 mbps. No communication is possible beyond 250 m. A simple opportunistic policy that gives an equal time share to each data rate is also in place. A max. of 5 packets are transmitted at 11 mbps, 3 packets at 5.5 mbps, and only 1 packet if the channel is good enough only for 2 mbps. Scheme 2 A plain IEEE A packet is sent at 11 mbps if the receivertransmitter distance is less than 100 m; otherwise no transmission.

74 Adaptive MAC and Plain IEEE Comparing Scheme 1 (adaptive-rate MAC) and Scheme 2 (plain IEEE ) for an 18-node linear topology Network

75 Conclusion The advantages of cross-layer design appear to be real and will be increasingly important in future. However, it will be useful to note some adverse possibilities and exercise appropriate caution in the CLD approach. Unbridled cross-layer design can lead to a spaghetti design and stifle further innovations since the number of new interactions introduced can be large. Every update may require complete redesign and replacement which can stifle proliferation. Cross-layer design proposals must therefore be holistic rather than fragmenting.

76 Thank You!