Research on qos assurance for multimedia data on wireless ad hoc network

of all competing nodes. Since no node fully controls its sending rate, the Kelly rate control al- gorithm can be translated directly to the contention window control, and new algorithms can be designed to achieve fairness and using the channel effectively. However, most current approaches focus only on a specific fair- ness level, not arbitrary levels, or allocate bandwidth by standard dis- tribution or weighted proportional. AOB, MFS, and Dynamic-802.11 mechanisims focus only on distributing bandwidth according to the standard distribution. IEEE 802.11e and P-MAC provide only weighted fairness. Some such algorithms (Dynamic-802.11, MFS, and P-MAC), also create network instability as they try to achieve channel utiliza- tion through dynamic CW tuning. This is because the solutions re- quire an individual node to run the iterative algorithms to estimate both the CWs used by the competing nodes and the number of compet- ing nodes. However, such estimation requires that all nodes, with or without the packet to transmit, must start simultaneously two dynamic tuning CW and iteration estimation algorithm, and run concurrently step-by-step these algorithms. Nodes with expired values of CW size 3 and number of competing nodes, due to asynchronous implementation or temporary errors (likely to occur with wireless networks), can cause these algorithms to fail. From the analyzed problems, the motivation of the thesis to re- search, improve the bandwidth control mechanisms. The aim is to further improve the QoS of wireless network services, especially with multimedia data. 3. Thesis objectives The general objective of the thesis is to research and propose solu- tions to improve the QoS of multimedia data services in ad hoc wire- less networks. Improved contention control mechanism in the net- work by dynamically adjusting the parameter value contention win- dow (CW) to achieve a flexible sharing ratio for different data types. Evaluate the effect of the QoS parameters on multimedia data types using the testbed system. 1. Study and propose methods to divide throughput according to the ratio of data streams with different priority in ad hoc net- works. 2. Study and propose method of CW dynamic control to achieve throughput according to the needs of data streams with different priority in ad hoc network. 3. Study and propose methods to evaluate wireless network service quality using testbed system. 4. Object and scope of the thesis QoS of multimedia data in wireless networks is a very large and 4 complex topic. To complete the research objectives, the thesis identi- fies the subject and scope as follows: a) Subject: Factors affecting QoS such as fairness index, total through- put, delay, packet loss for multimedia data. b) Scope: The solution to ensure quality of service for multime- dia data (voice, video, background) on fairness index and total throughput. 5. Research Methodology To accomplish the proposed research objectives, the thesis uses re- search methods: theoretical, simulation, and experimental evaluation (testbed), specifically as follows: a) Theoretical: dissertation research and synthesis of works related to the service quality of wireless ad hoc networks published at home and abroad. Also, the thesis focuses on analyzing in detail the pros and cons to detect the existing problems in order to propose suitable improvement solutions. b) Simulation: thesis evaluated the results by simulation using NS2 software. For the results to be objective, the thesis simulates many different scenarios with many different parameters, based on the sim- ulation results, the thesis evaluates the achieved efficiency in terms of equity index and total throughput. Make a statement of the advantages and disadvantages of the proposed solution. c) Testbed: the thesis uses the evaluation method based on the experimental system (testbed) to achieve the practicality because this method uses real devices, software to generate real data, so the results will be closer to reality. 5 6. Thesis layout The thesis is composed of three parts, namely “Introduction”, “Con- tent” and “Conclusion”. In which the most important part is the “Con- tent” includes the following chapters: • Chapter 1: “General research” research and analyze main prob- lems of the thesis and available solutions. • Chapter 2: “Throughput Analysis of IEEE 802.11 networks ” performs the calculation of the maximum theoretical through- put of IEEE 802.11 wireless networks. Of the factors that influ- ence QoS, throughput plays an important role. However, with the IEEE 802.11 wireless standard, throughput is a factor that has many different “values”, for example, with 802.11b hav- ing a throughput of 11 Mbps but this is the data rate of radio waves (radio data rate) rather than the rate at which the pack- ets are transmitted (the primary factor in network throughput). Theoretical maximum throughput is important because it can be used to provide the network at the optimum level for data trans- mission, especially for multimedia data. The 802.11 family of standards includes many technologies such as 802.11a/b/g/n/ac. This chapter of the thesis performs theoretical computation with two popular standards 802.11b and 802.11g. • Chapter 3: “Proposed method of controlling data flows with different priority ” performs simulation-based evaluation to prove that despite IEEE 802.11 can provide bandwidth splitting for different types of multimedia data, but this does not really guar- antee the service quality requirement of the data streams. For example the voice type data always gets the highest rate and 6 the background data is always the lowest. Therefore, in some cases such as with real-time data where service difference is re- quired for best effort and real-time variable data traffic, IEEE 802.11 does not. can provide QoS commensurate with such a request. And it is therefore necessary to have a more flexible di- vision mechanism. In the framework of the thesis, the problem is solved by measuring real data at each node receiving data for a period of time, and then comparing with the theoretical data to determine whether to increase or decrease the value. Contention Window – CW, the proposed algorithm will control the increase or decrease of this CW value to achieve a flexible split ratio that suits the needs of users with multiple Different data types such as voice, video and background, the evaluation of the proposal is performed on the NS-2 network emulator under many scenarios. • Chapter 4 “Proposed method of evaluating flux control solution by testbed ” analyzed the advantages of the experimental-based evaluation method (testbed), then built Building an evaluation system using testbed, proposing evaluation steps, and then us- ing it to evaluate the effect of changing the value of the set of parameters of multimedia data service quality affecting factors such as throughput, latency, and packet loss rate in the multi- stage ad hoc model, the results prove that using real hardware devices will have results closer to reality instead of ideal values. in theoretical modeling, simulation or analysis methods. 7. Contribution of the thesis Through synthesizing and analyzing the advantages and disadvan- tages of previous studies, the thesis proposes a number of solutions to 7 promote the advantages and overcome the limitations. Here are some of the main contributions of the thesis: 1. Thesis has proposed a method to improve the share of bandwidth to achieve a fair level of multimedia data.[CB1] 2. Thesis has proposed a method of controlling data flows with dif- ferent priorities to achieve fairness and maintain high through- put when the network is in saturation.[CB4][CB6][CB10] 3. Thesis proposes a method to evaluate wireless network perfor- mance by testbed system. [CB7][CB8][CB9] From the results achieved, the thesis finds that the quality assur- ance of multimedia data services in the ad hoc wireless network is complex. The approach based on dynamically controlling the con- current window value as well as changing the set of service quality parameters evaluated by the emulator and experimental system has shown efficiency and potential for application to solving. solve the problem of multimedia data service quality in ad hoc networks. 8 CHAPTER 1. LITERATURE REVIEW 1.1. Introduction to IEEE 802.11e In 1997, Institute of Electrical and Electronics Engineers – IEEE created the first Wireless LAN standard (WLAN), which is the 802.11 standard. The IEEE 802.11 standards have gone through a long evolu- tionary history. Among the many 802.11x standards, the IEEE 802.11e proposed in 2005 is notable for offering a set of Quality of Service (QoS) focused on multilateral applications. This standard, such as voice, video, and IEEE 802.11e, was incorporated as a part of the IEEE 802.11 WLAN standard in 2012. The highlight of IEEE 802.11e is that it controls textit access to distributed channels. Enhanced Distributed Channel Access (EDCA) is based on a contention-based mechanism, which is consistent with the distributed characteristics of the nodes in ad hoc wireless networks. EDCA mechanism uses a differentiated medium access method, using different priorities for each type of data stream. EDCA defines four priorities type under Access Categories – AC for different data types and has distinct services for each of these AC types. Whether different data frames are mapped to ACs will depend on the upper layer service quality requirements. Each frame from the upper layer to the MAC layer is weighted User Priority – UP depending on the application that generated the frame. There are eight priority weight values as shown in Table 1.1.1. The EDCA mechanism handles the competition of transmission access based on the following parameters: Arbitrary InterFrame Space Number – AIFSN, is the number of time-slots (SlotTime) after each SIFS time period that a station must wait before entering a reverse or 9 Table 1.1.1. User Priority and Access Category Priority UP AC Data type lowest 1 AC BK Background - 2 AC BK Background - 0 AC BE Best effort - 3 AC BE Best effort - 4 AC VI Video - 5 AC VI Video - 6 AC VO Voice highest 7 AC VO Voice Table 1.1.2. Default value of EDCA paramemters Parameter BK BE VI VO AIFS 7 3 2 2 CWmin 15 15 7 3 CWmax 1023 1023 15 7 TXOPLimit (ms) 0 0 1.504 3.008 data transmission phase; Contention Window — CW: each station cal- culates the total backoff time value from a random value taken within the limit of the window size; TXOP limit, is the maximum transmis- sion time per station after a Transmission Opportunity has been won. In general, AC with higher priority will have AIFSN, CWmin, CWmax smaller and T XOPlimit larger than AC with lower priority. These EDCA parameters are different for each AC type, and are detailed in Table 1.1.2. Corresponding to the parameter sets for these priorities, network performance parameters such as delay, packet loss and especially through- 10 put there are also differences between data types. 1.2. Conclusion In this chapter, the dissertation explores relevant studies on service quality in wireless networks, especially for multimedia data. The the- sis explores the IEEE 802.11e standard as a proposal to focus on ensur- ing quality of service for multimedia data, which is currently accepted as a part of the IEEE 802.11 family of wireless standards. The study also raised several issues affecting wireless QoS and several methods of evaluating wireless network performance. 11 CHAPTER 2. THROUGHPUT ANALYSIS OF IEEE 802.11 WIRELESS NETWORK This part of the thesis will calculate the maximum theoretical through- put of 802.11 b, g wireless networks. Theoretical maximum through- put is important because it can be used to provide the network at the optimum level for data transmission, especially for multimedia data. 2.1. IEEE 802.11 theoretical throughput analysis To calculate the theoretical throughput, we have the following for- mula: T hroughput(Mb/s) = Amount of data (bits) Transmission time (s) (2.1.1) 2.1.1. IEEE 802.11b By calculation, the thesis obtained the results in Table 2.1.1 on the theoretical throughput of IEEE 802.11b. Table 2.1.1. Mean theoretical throughputs (backoff counters of 15.5 and 0) for 802.11b Layer Payload Backoff 15.5 Backoff 0 Speed (Mb/s) Speed (Mb/s) 1 5.5 11 1 5.5 11 Application 1470 B 0.91 4.18 6.97 0.93 4.70 8.54 UDP 1478 B 0.91 4.20 7.01 0.94 4.72 8.58 IP 1498 B 0.93 4.26 7.10 0.95 4.79 8.70 LLC 1506 B 0.93 4.28 7.14 0.95 4.81 8.74 12 Table 2.1.2. Mean theoretical throughputs (backoff counters of 15.5 and 0) for 802.11g Layer Payload Backoff 15.5 Backoff 0 Speed (Mb/s) Speed (Mb/s) 6 12 36 54 6 12 36 54 APP 1470 B 5.09 9.26 20.38 25.48 5.42 10.40 26.88 36.52 UDP 1478 B 5.12 9.31 20.49 25.62 5.45 10.46 27.03 36.72 IP 1498 B 5.19 9.44 20.77 25.97 5.52 10.60 27.39 37.22 LLC 1506 B 5.22 9.49 20.88 26.11 5.55 10.66 27.54 37.42 2.1.2. IEEE 802.11g By calculation, the thesis obtained the results in Table 2.1.2 on the theoretical throughput of IEEE 802.11g. 2.2. Comment on theoretical calculations Let us first compare the theoretical results obtained in the 2.1.1 and 2.1.2 sections for efficiency. See Table 2.2.1 and 2.2.2 show that the efficiency is higher at lower speeds. This happens because with a fixed amount of data, at a lower rate it takes more time to send, reducing the impact of time charges such as DIFS, backoff procedures or PHY header on throughput. Table 2.2.1. The effect of theoretical throughput at the Application layer of 802.11b Speed (Mb/s) 1 5.5 11 91% 76% 63.4% Next we calculate the percentage difference between the nominal value and the actual value of throughput at the same rate. For example, the difference between 5.5 and 6 Mbps (nominal speed) of 8.33 % is 13 Table 2.2.2. The effect of theoretical throughput at the Application layer of 802.11bg Speed (Mb/s) 6 12 36 54 84.83% 77.17% 56.61% 47.19% calculated using the formula 2.2.1. Di f f erence(%) = (1− 5.5Mbps 6Mbps )×100 = 8.33% (2.2.1) If we calculate the difference between their respective throughput, we get 17.88 %. This means that the standard 802.11g protocol is 17.88 % more efficient in throughput than 802.11b. And when com- pared to the rate of 11 and 12 (Mbps): the difference between the nominal speed is still 8.33 %, but the difference in throughput is al- ready 24.73 % .If we used mode b/g mixed, the g clients will lose a lot of performance. So, knowing for sure that there are only g clients in the network, it is better to disable speed b on the Access Point. 2.3. Conclusion Through the above calculations it can be seen that the throughput value depends on many factors and may have many different values depending on the communication conditions. Determining the max- imum theoretical throughput will help assess the network throughput according to more practical solutions such as simulation or experiment with an additional measure to compare and compare performance re- sults. 14 CHAPTER 3. PROPOSE METHOD FOR CONTROLLING PRIORITY-BASED NETWORK FLOWS 3.1. A method to improve throughput sharing ratio for priority- based data flows 3.1.1. Proposed method of sharing throughput by ratio for data flows with different priorities The QoS function in IEEE-802.11 provides a priority for each type of traffic, but it cannot guarantee the right throughput rate for each traf- fic type. The thesis proposes a weight for each priority level different from the default IEEE-802.11 QoS settings, these numbers will cor- respond to the throughput rate in the proposed solution. That means, with a suggested number of 3:2:1:1, the proposed method will achieve a throughput share rate [total of 2 Mbps] in Figure 3.1.1. Figure 3.1.1. Weighted sharing throughput with expected ratio. To assess whether the share of throughput is achieved according to the desired proposed rate, the thesis uses the fairness index, as de- fined by R. Jain, based on which the thesis propose the formula ?? to calculate the weighted fairness as follows to evaluate the level of 15 proportional flux distribution: FairnessIndex = ( n ∑ i=1 xi ki )2 n× n ∑ i=1 (xiki )2 (3.1.1) Here, n is the number of threads, xi is the end-to-end throughput of flow i, and ki is the weight corresponding to the data types. The results will range from 1/n (worst) to 1 (best), and maximized when all flows receive the same channel allocation. This fairness value will be used to evaluate the bandwidth sharing rate of different data types, ie the closer to one (1) the closer the desired ratio is to be achieved. Fairness Index is evaluated based on effective throughput (goodput) at the target station. 3.1.2. Proposed method of dynamic CW control to achieve on-demand throughput of data flows with different priorities Based on the cross-layer scheme to ensure fairness in IEEE 802.11, we propose to improve MAC stratification in IEEE-802.11 to achieve fairness between different data streams (e.g. video, voice, text,. . . ). We have done this based on two modules named TX Flow Estimation and Utilization Estimation. Module TX Flow Estimation works at MAC sub-layer to count the number of flows in the transmission range. We call these flows are TX flows. A flow TX flow is determined by the MAC address, the IP address of the source and destination, and the flow type (eg Voice, Video, Best-effort), by decoding the packet’s header. We define TX flows as nT X [i] where i stands for {Voice, Video, BestEffort}. Note that in this case, TX flow is end-to-end (end-to-end or host-to-host) 16 data streams and it may include some flow between threads. process (process-to-process). Assume that there are n data streams where ki is the weights of the QoS data types defined in IEEE-802.11. For example, the throughput ratio of Voice, Video and Best Effort is 3: 2: 1, we have kVoice = 3, kVideo = 2, kBestE f f ort = 1. Next we define the fair share of bandwidth for each stream using the formula: Fair Share Ratio[i] = ki n ∑ i=1 ki×nT X [i] (3.1.2) Module Utilization Estimation evaluates the actual link perfor- mance of the flow. The link performance is determined by the analysis interval (observation period) Active T ime[i] of the flow over a prede- fined time period EP. The stream Active T ime[i] time is defined as the time spent transmitting packets in the stream i. The algorithm 3.1.1 is used to estimate Active Time[i] value of data stream i by sending packets. We take the time to send the packet during a given time EP equal to eighty percent of the current sending time plus twenty percent of the time previously sent as described in Algorithm 3.1.1 below. 17 Algorithm 3.1.1: (Active Time[i]) Initialization: Active Time[i] = 0 TActive[i] = 0 for each each interval time EP Active Time[i] = 0.8×Active Time[i]+0.2×TActive[i] TActive[i] = 0 for each each packet p if p→ destID == localID if p→ Type ==CT S TActive[i] = TActive[i]+TRT S+TCT S if p→ Type == ACK TActive[i] = TActive[i]+TDATA+TACK +TBacko f f The value Real Share Ratio[i] is defined as the ratio of Active Time[i] to the estimated time EP as below formula: Real Share Ratio[i] = Active Time[i] EP (3.1.3) In this formula 3.1.3, the estimated time (EP) is the given obser- vation time, the value is chosen so that it is not too short to ensure the observation. being continuous as well as not too long will distort the simulation, so when doing the simulation with NS-2, the value of EP is chosen to be two seconds (2s). CW will be adjusted, and throughput will be assessed during this EP (value) process. The corrected CW value will be determined by the formula: 18 CW ′[i] = Real Share Ratio[i] Fair Share Ratio[i] ×CW [i] (3.1.4) In the above formula, the equity value Fair Share Ratio[i] is used as a threshold for the channel access priority. If threads find that the actual Real Share Ratio[i] is less than their Fair Share Ratio[i] share then the thread will use a smaller CW value in phase back-off. As a result, you can increase channel access opportunities, which means higher bandwidth is allocated. Conversely, when the thread realizes that Real Share Ratio[i] is greater than its Fair Share Ratio[i] share then the thread will use a larger CW value in the phase. backtrack- ing. That will reduce your chances of accessing the channel, leading to other streams having more chances to access the channel. In the case of streams with only a small recommended load, they will access the channel more easily, and the remaining bandwidth will be shared among other threads. This will allow for more efficient use of channel bandwidth and ensure fair bandwidth sharing between streams. 3.2. Conclusion This part of the thesis proposes a mechanism to improve the rea- sonable bandwidth sharing between streams by adjusting the CW in a more reasonable way, this solution provides a flexible CW value, cor- responding to the prioritization of throughput for different data types to achieve a reasonable level of sharing between multimedia flows. The results evaluated and simulated by NS-2 demonstrated that our method is better than the original IEEE-802.11e standard in some simulation models. The above research results are shown in [CB1], [CB2], [CB3], [CB4], [CB6] publications. 19 CHAPTER 4. PROPOSE METHOD FOR CONTROLLING THROUGHPUT BY TESTBED SYSTEM 4.1. Evaluating wireless networks using testbed The thesis proposes a flowchart as shown in Figure 4.1.1 which shows the whole process of using testbed to analyze and evaluate net- work parameters. Figure 4.1.1. Experimental research method. 4.1.1. Evaluate the effect of QoS parameters Sender node Receiver nodeForwarding node Figure 4.1.2. Multi-hop ad hoc network. 20 Figure 4.1.2 is a testbed included two connected wireless nodes in ad hoc mode via an forwarding node, these three nodes support QoS for multimedia data (WiFi Multimedia or WMM feature) with IEEE 802.11g. This ad hoc topology is setup by hostapd [?] with WMM turned on. The default QoS parameters of IEEE 802.11g, with four Access Category types with priority level from lowest to high- est: AC BK (background), AC BE (best effort), AC VI (video) and AC VO (voice) with corresponding IEEE 802.11 QoS [?] parameters set, are shown in Table 4.1.1: Table 4.1.1. Default QoS parameters. AC CWmin CWmax AIFSN TXOP limit (ms) AC BK 15 1023 7 0 AC BE 15 1023 3 0 AC VI 7 15 2 3.008 AC VO 3 7 2 1.504 Effect of Contention Window (CW) Next, we demonstrate the relationship between Contention Win- dow size of three data types (Voice, Video, and Background). To do that, we fixed two parameter sets of the highest (voice data) and the lowest (background data) priorities and step-by-step changing CW- size of video data. The range of CW values of three types of data which also change from arcording to Table 4.1.1 and CWmin size of the video data which will change in that range (3 to 15) to give an observation of how the throughput ratio of three data types changes corresponding to CW changes. Looking at the experiment result in Figure 4.1.3, we see that the 21 CWmin size greatly affects network throughput, small CW has large throughput, and vice versa. With default CWmin (7), the video through- put is still as large as the default priority. But when the CWmin in- creases larger, the throughput of video data immediately dropped very quickly, as well as jitter and packet loss ratio become much worse. (a) Throughput (b) Jitter (c) Packet loss Figure 4.1.3. Effect of CW parameter Effect of TXOP The effect of Transmission Opportunity (TXOP) limit is simple. If this value gets larger, the throughput of corresponding node becomes larger, too. Otherwise, the jitter (or delay in average) for other nodes would be increased because these nodes must wait for a longer trans- mission time to send their packets. Because of the largest TXOPlimit value (3008 microseconds) as default in IEEE 802.11. We observe the change of network performance by changing the TXOP values of voice instead of video data. As shown in Figure 4.1.4, if the TXOP value of voice data gets larger, the corresponding throughput becomes larger. But when the TXOP of voice increases to its deafault value (47 in Figure 4.1.4 or 1504 microseconds in Table 4.1.1), the throughput is still increasing but at a fairly small level. 22 (a) Throughput (b) Jitter (c) Packet loss Figure 4.1.4. Effect of TXOP parameter Figures 4.1.4 indicate that TXOP does not affect the jitter and packet loss ratio val