ENSC 427- Communication Networks Spring Quality of service of the wireless networking standard over a Multi user environment.

Transcription

1 ENSC 427- Communication Networks Spring 2015 Quality of service of the wireless networking standard over a Multi user environment Group 9 Name: Saumya Sangal ssangal@sfu.ca Name: Jasmine Liu zyl2@sfu.ca

2 Page 1 Table of Contents 1. Abstract Introduction... 2 a) What is Wi-Fi?... 2 b) What are the wireless standards?... 2 c) What is QoS?... 3 d) Some of the network performance parameters Riverbed Simulation Model... 4 Part A: Application Definition... 5 Part B: Nodes Definition Riverbed Simulation Analysis and Discussion...13 QoS parameter analysis...14 Case 1: Introduction of streaming device...16 Case 2: Introduction of Gaming device...22 Case 3: Introduction of VoIP device...27 Case 4: Introduction of low bandwidth users Supplemental Analysis Results Future Work References...38

3 Page 2 1. Abstract Wi-Fi is a term synonymous with daily life as we utilize this technology as a basis for a variety of activities. Hence, a constant question remains as to how to increase the speed and efficiency of our Wi-Fi network. In our project, we seek to analyse, compare, and scrutinise the wireless networking standards in terms of the quality of service they provide to the users. For this purpose, we shall consider a multi user environment consisting of mobile users and shall study the effects of an increase in the number of users as well as a variation in distance from the access point. Our analysis shall take into account throughput and delay as we seek to complete our objective. 2. Introduction a) What is Wi-Fi? Wi-Fi is a wireless networking technology that provides a connection between a sender and receiver using Radio Frequency technology (RF) [9]. An example of a radio frequency band is the frequency of 5GHz used to provide high speed internet connections and was introduced by Wi-Fi Alliance [1]. Wi-Fi is wireless, which means there are no physical connections between sender and receiver, this is ideal for home, school, office, and any public area to reduce the wiring as well as provide mobility for users. RF technology is built on the principle of electromagnetic fields propagating in space. The field is produced by antenna s which convert RF current to the EM field [9]. One of the most fundamental components of a Wi-Fi setup is an Access Point (AP) whose prime responsibility is to broadcast the wireless signal [9] which is then picked up by devices within range. In order for a device to support the use of Wi-Fi it must be declared Wi-Fi certified by the Wi-Fi alliance [9]. This certification acts as a validation of the device being interoperable with any other Wi-Fi capable device regardless of its make or manufacturer. This allows a device to access and use any access point and vice-versa. b) What are the wireless standards? Wi-Fi is based on IEEE standard protocol called [1]. The standards have a, b, n, g, ac protocols, for this project we will implement using the n protocol n works in the 2.4 GHz frequency band and uses orthogonal frequency-division multiplexing (OFDM) transmission scheme instead of Complementary Code Keying (CCK)[2]. The standard has a max data rate of 600Mbps and a range of approximately 70 meters indoors, this is satisfactory for our simulation at 26 Mbps for a small environment.

4 Page 3 Table 2.1: comparison between different wireless standards [14] c) What is QoS? QoS is a measurement of the network performance seen by users. Quality of Service (QoS) for networks is an industry set of standards for ensuring high-quality performance of applications [3]. The concepts of QoS is all network traffic are receiving the best effort delivery service. For example, if one application requires a lot of bandwidth, it will affect the performance of all other applications in the same network. Therefore, the goal for QoS is to provide the applications a sufficient bandwidth for its purpose with minimal probability of jitter, latency, and data lose, hence improves the reliability of network. d) Some of the network performance parameters Throughput: The rate of successfully transfer the data over a communication channel. [3] Errors: The message is corrupted due to bit error during the transmission period. [3] Latency: The time delay of transmitting and receiving the message from source to destination.[3] Jitter: The variation of latency [3]

5 Page 4 3. Riverbed Simulation Model Our simulation aims to create a relevant and real life scenario incorporating situations which occur on a highly consistent basis. The aim is to examine the immediate and short term real life effects on the quality of service for a variety of users when presented with a situation where each user is connected to the same router/access point. For the purpose of this report we take a sample size of 100x100 meters. We intend to visualize such a scenario with a number of fixed users located within the confines of our 100x100 sample space using the available wireless connection on their laptops/tablets/smartphones for a number of different applications that we shall define further on. We then introduce new mobile users to this environment of 4 established and fixed users. Our study now governs two basic questions: Q1: What is the effect that the new user has on the quality of service to the existing users with our sample space? It is worth noting here that in each case we present we introduce a user using an application that varies in its level of throughput to the existing users in some way regardless of it being the same nature. It is for this purpose that we define a total of 8 applications. Q2: What is the effect on the quality of service to the existing users if these new mobile users now move around the sample space during our simulation? Note for question two we wish to compare the changes between the situations where the new user is stationary and when the user is moving in the sample space. For this purpose we made purpose built trajectories and assigned the new users to transverse along them. We shall examine each case from the perspective of the n 2.4GHz wireless standard at a rate of 26Mbps. This rate is our default value and affords us the chance to answer a third question. Q3: What is the Delay and throughput for our given model according to the various wireless standards? In answering the question we can examine how our simulation model would have compared had we chosen a different operating environment and thus help us conclude if the standard and data rate we have chosen are ideal.

6 Page 5 Part A: Application Definition The applications perform as per a specified type of service that allocates a level of priority to the application. These levels are represented in the diagram below. There are a total of 8 applications in our simulation model varying in the level of throughput and delay. Table 3.1: Level of priority table [7] Figure 3.1: Applications in our simulation model 1. VoIP The first application we introduced in our analysis was a simple VoIP application. Considering the high number of tablet and smartphones that we implement in our simulation model it was imperative to have a high quality VoIP application. The application was given priority level 6 of interactive voice thus indicated extremely high priority in our model. The speech quality was chosen to be GSM quality speech.

7 Page 6 Figure 3.2: VoIP attribute 2. Http The second application took into consideration the most common form of Wi-Fi use. The application was set to Heavy web browsing at a standard rate with a priority of 2 in our model. Figure 3.3: Http attribute 3. YouTube High Resolution Video After having accounted for voice over IP and web browsing application we aim to implement streaming multimedia in our model. For this purpose we name our application YouTube as a symbol of its representation of a video stream. In this case we strive for a high resolution video stream that requires more bandwidth usage. The stream is implemented through the video conferencing mode where we choose the quality to be High resolution and then further make edits to make the frame interarrival time as 30 frames per second. The video size is set to 352*240 pixels and we set the outgoing stream to NONE as we do not wish to send data out for our application. We set a priority of 4 corresponding to streaming multimedia.

8 Page 7 Figure 3.4: YouTube high resolution attribute Figure 3.5: YouTube high resolution frame 4. YouTube Low Resolution Video We also accommodate an application that is built on a low resolution video platform. We repeat the same steps from the earlier streaming application however we reduce the quality of the video was assigning an interarrival time of just 10 frames per second. In our model we set this to be represented by a mobile smartphone/tablet user. The priority is the same in this case. Figure 3.6: YouTube low resolution attribute

9 Page We also take into account a situation where a user may be checking their within our sample space. The level of usage is set to medium load at a priority level of standard [2]. Figure 3.7: attribute 6. Google (Search Engine Usage) The sixth application is set to represent a user who is searching information online. This application can be chosen from the drop down list on the http application tab. The searching feature was set to be best effort (0) as a priority level. Figure 3.8: Search engine attribute 7. Game-Counter Strike The game application is the first of the two custom made application that are set to run on our simulation model. The game is set to run as FPS- First person shooter game. The game was chosen as it requires extremely fast response times from the gamer resulting in frequent clicks that can be analyzed in our simulation model. For this purpose we have sought to recreate one of the most popular FPS games in Counter Strike.

10 Page 9 Our model is inspired and based on the aggregate values examined by group 3 from spring 2010 and group 4 from The values are listed in the study done by Johannes Färber, Network Game Traffic Modelling,which describes the statistics involved in a 36 hour capture with 50 participants [15] These values can be seen below: Server client interarrival time (ms) extreme (55,6) constant (40) packet size (bytes) extreme (120,36) extreme (80, 5.7) Table 3.2: Server and Client interarrival and packet size comparison In order to implement this game we began by creating a new application called Game in our application profile. Here we choose to make our own custom application instead of the usual case of choosing a pre-defined application. Now within that we choose task description and add game with weight 20. The task ordering remains as serial. Now we choose our type of service as Interactive Multimedia with a priority of level 5. Figure 3.9: Gaming configuration Now that our application has been made we define the profile for the said task and name it gameplay. To implement this game we must now add a task profile in order to instruct the flow of traffic for our game model. On creation of a new task called game task we change the manual configuration policy by inputting out packet size and interarrival time as listed in the table above. Figure 3.10: Gaming traffic configuration

11 Page 10 Now as a final step we must change the HCF parameters of our gaming node in order to allow. HCF is indicative of High cycle fatigue and may result in huge delays for our gamer due to the other applications we will run concurrently. Figure 3.11: Gaming attribute 8. Skype For our last application we create the Skype application. In order to do so we need to capture the appropriate packet data in order to create our own custom application. To do so we downloaded Wireshark. Within Wireshark there exists a large database of pre-existing packet files to examine. For our purpose we downloaded the SkypeIRC.cap file. The file can be seen below where we have chosen traffic flow in and out in regards to a single port ( ) to model as our default case. The image below is separated in two columns representing the packet arrival time and the packet length. The data is then exported as a text file to Microsoft excel so we may make aggregations in order to arrive at the mean outcomes. Figure 3.12: Skype packet capture

12 Page 11 Using excel we arrive at the mean outcomes for the interarrival time and packet size by using the data for length(bytes) and arrival time(ms)now to input our data for skype we create a new application and name it skype. Within skip we choose the video conference option and edit it. The mean values obtained from excel and csv file is input to create a working skype application. Within our video conference feature we ensure that the flow of bytes and frames is set to equal both ways. We choose an exponential function due to the exponential nature of our mean in excel. Figure 3.13: Skype frame interarrival and frame size Our skype application is then allocated interactive voice type of service with priority 6.The skype application is classified under a profile of the same namesake. Figure 3.14: Skype configuration attribute

13 Page 12 Part B: Nodes Definition Nodes Functionality Allows us to define all our application and their level of service and priority. Our model defines 8 such applications as per our need. Allows us to classify applications so as they may be available to wireless devices to use. The task config is defined for our Counter Strike game to define the packet size and interarrival time. The android is used in multiple situations especially implementing our Skype application. The tablet is used in multiple applications including in our trajectory case studies. Supports the profiles and applications via an Ethernet link to access point. Used for multiple applications as a new user device. Links the server to the router.

14 Page 13 Acts as a stationary wireless module in our model Functions as an access point for our wireless devices Table 3.3: Nodes definition in our simulation model 4. Riverbed Simulation Analysis and Discussion In our simulation we begin by running each scenario/case at the predefined standard of n 2.4GHzat 26Mbps. We have 4 wireless mobile nodes represented in our sample space. These nodes are mandated to be the 4 stationary cornerstones of our analysis and each of these nodes does not move from its position so that we may better gauge the effect of additional users. The nodes represent the following cases where each user has a different effect on the quality of service of the other. The simulations were run were for a period of roughly 5min throughout our analysis as our Riverbed academic model was limited to roughly events and since we utilize high throughput devices we had to settle for a mean time that suited each case we present. Our 4 main stationary devices are: Device Application Priority level android Skype interactive voice(6) laptop (mobile workstation) YouTube streaming multimedia(4) laptop Counterstrike game interactive multimedia(5) laptop Heavy web browsing standard (2) Table 4.1: Stationary devices in our simulation model

15 Page 14 Figure 4.1: Main sample workspace QoS parameter analysis a. Throughput For our basic model sample space, an analysis of the throughput at 26Mbps is displayed in the graphs to follow. We can see that there is large amount of spiking in our graphs for the skype, counter strike and YouTube user. This can attributed to the fact that we have three high demand applications running together. Furthermore our simulation does not run for a long period allowing these applications to settle and thus we have a burst nature for our simulation. We notice that the highest throughput is for our android device using the skype application. This would make sense since the Application is very demanding on a network and we have simulated using real life values instead of the pre-defined stable cases. The next highest throughput is for our YouTube high resolution video user which again would make sense since the video is incoming at 30 fps. The constant burst nature is attributed to the user clicking while playing the game. This is in line with our expectations as we chose the FPS game due its high input responsiveness from the user end. The web browser is the least demanding and we can see it maintains the same throughput rate throughout our simulation. There is no set up or buffer time for out browsing and thus it starts well before the rest. We see the spikey nature corresponds to the user frequently clicking on different links while browsing. Overall we see there is clear correlation between the priority levels and the throughput of our users. b. Delay In terms of delay we see that our high quality video applications for the YouTube user and the skype user are significantly high. The delay for the web browser is the lowest since it is not a very demanding application. However there is a huge surge web browsing delay towards the end of the simulation as the modeler assigns the other three applications to finish their task

16 Page 15 simulation and since web browsing is at the lowest priority it must suffer a large amount of delay. Figure 4.2: Android- skype throughput Figure 4.3: gamer Figure 4.4: YouTube user Figure 4.5: Web browser

17 Page 16 Figure 4.6: Overall throughput Figure 4.7: Overall delay Case 1: Introduction of streaming device Part A: Stationary new user We now introduce a low resolution streaming tablet within the confines of our simulation. In this scenario the additional node is set to be stationary within the sample space. The node is set to run a low resolution streaming application. Figure 4.7: Case 1 layout in workspace

18 Page 17 Qos parameter: a. Throughput On introducing a new user who uses a tablet to stream a low resolution video we observe a large decline in the throughput of our video applications. The skype user experiences a 90% decline in the throughput. However the throughput of the gamer increases equally drastically. The web browsing throughput remains virtually the same however there is sharp decline towards the end of our simulation. The YouTube user now experiences more spiking now in the quality of the video being experienced. As evidenced we see that both the low resolution and high resolution user experience a large amount of spiking whereas the throughput of the user was nearly constant in our basic sample model. b. Delay In terms of delay there is significant delay in each of the streaming video applications and our web browsing application now has a spike in terms of a delay much sooner than our basic modeling case. The delay for the other two remain roughly the same. Figure 4.8: android skype Figure 4.9: game throughput

19 Page 18 Figure 4.10: web browser Figure 4.11: low res video Figure 4.12: smartphone YouTube

20 Page 19 Figure 4.13: Throughput Figure 4.14: Delay Part B: Trajectory defined for new user We now define a trajectory 1 for our new user to move through. The projection of the trajectory can be seen in the sample space diagram below while the trajectory is configured to move across the sample space at a constant rate throughout the simulation. In order to do so we define the transverse time as 85 seconds for each of the 4 direction changes the user makes. It is imperative to note that we use smartphone and tablet users to do so as such a scenario is very likely in any situation these days. Figure 4.15: Trajectory information defined for new user

21 Page 20 Figure 4.16: New user introduced workspace layout Qos parameter: a. Throughput With the new user now moving we compare the throughput with that of the stationary case. There is sharp improvement in the throughput of the skype application corresponding with a decline in the gaming application throughput. However the key observation here is that the throughput for the low resolution video user increases drastically as the user moves closer to the router/access point. The effect of this increase is reflected in the throughput for our web browser who experiences low quality of service. b. Delay In terms of delay we observe that there is a decrease in delay for the low quality video user and at the same time there is an increase in delay for our skype user and gamer. Further since the web browser user does have any throughput for this duration we do not see much delay however towards the end there is an increase in delay indicating that had the simulation run longer we would have seen throughput for the user come through.

22 Page 21 Figure 4.17: android skype throughput Figure 4.18: game throughput Figure 4.19: smart YouTube throughput Figure 4.20: low res video

23 Page 22 Figure 4.20: http browser Figure 4.21: overall throughput Figure 4.22: overall delay Case 2: Introduction of Gaming device Part a: stationary Qos parameter: a. Throughput In this scenario we introduce a stationary tablet user to our sample space. The tablet user utilizes the same gaming application that our laptop user in the basic model uses. The throughput for the skype user takes a massive fall however the throughput to the gamer increases drastically. At the same time the throughput for the YouTube user and the http browser. It is interesting to observe that while the laptop gamer has increased

24 Page 23 throughput the same cannot be said for the tablet gamer who experiences sharp high and lows such that the game is almost unplayable at a very low throughput. b. Delay In terms of delay no significant change was observed from our basic sample model case. Figure 4.23: android- skype throughput Figure 4.24: web browser Figure 4.25: game through Figure 4.26: smart YouTube

25 Page 24 Figure 4.26: tablet gamer Figure 4.27: overall throughput Figure 4.28: overall delay

26 Page 25 Part b: trajectory defined The tablet is set to follow the same trajectory 1 for our analysis. Qos parameter: a. Throughput With the tablet now moving the throughput for most of our application remains more or less the same indicating that there is not heavy effect on them due to the varying movement of the tablet gamer. What is important to note here that there is no change in the throughput for our tablet gamer but instead the throughput to the YouTube user increases significantly as a result of the movement of the tablet user. b. Delay In terms of delay there is a massive increase in the delay of our web browser user due to the movement of the tablet gamer. Figure 4.29: overall delay Figure 4.30: android-skype throughput Figure 4.31: gamer delay

27 Page 26 Figure 4.32: web browser throughput Figure 4.33: smart YouTube Figure 4.33: tablet gamer throughput

28 Page 27 Figure 4.34: overall throughput Figure 4.35: overall delay Case 3: Introduction of VoIP device We now introduce an android device using VoIP application set at interactive voice type of service with a priority level of 6. Part a: stationary VoIP user Qos parameter: a. Throughput On adding a standard VoIP user (not the skype user) we see that the VoIP throughput is pretty consistent at a level we expect it to be since our default VoIP simulation was stable at around bits/sec. We can observe that our gamer experiences a sharp rise in throughput. At the same time the skype user experiences a fall in the throughput experienced. The YouTube user and the web browser user both have slight changes. b. Delay In terms of delay we see that there is decrease in the delay of the web browser user accounting for a far more linear and lower overall delay.

29 Page 28 Figure 4.36: android VoIP Figure 4.37: game throughput Figure 4.38: web browsing throughput Figure 4.39: smart phone YouTube

30 Page 29 Figure 4.40: android skype through Figure 4.41: overall throughput Figure 4.42: overall delay

31 Page 30 Part b: Trajectory defined for the VoIP user Qos parameter: a.throughput The throughput for the android device using a VoIP application remains the same while moving around our sample space. The skype throughput and the web browser throughput falls through significantly. At the same time the gamer throughput increases significantly and the YouTube user increases a small burst of increased throughput. b. Delay The web browser user experiences sharp step increases at regular incremental time levels. Figure 4.43: trajectory information for the VoIP user

32 Page 31 Figure 4.44: android- VoIP Figure 4.45: android - skype Figure 4.44: game throughput Figure 4.45: web-browsing

33 Page 32 Figure 4.45: smartphone YouTube Figure 4.46: overall throughput Figure 4.47: overall delay

34 Page 33 Case 4: Introduction of low bandwidth users In this case we now introduce 3 additional mobile users each using low demand applications. The applications represented are , Searching and an additional web browser. Since they are low demand we do not expect too many variations in the quality of service and thus study the effects as one case. Qos parameter: a. throughput In our case for 3 additional nodes we do not experience much change in the throughput of our applications in both the stationary and the moving along trajectory case. This can be attributed to the fact that all 3 additional users are very low throughput in nature. b. Delay In the stationary case we see that the user experiences large delays in reaching its destination. In our trajectory case we see that each of the low throughput users experience significant delays while the high throughput users suffer no additional delay. It must be noted here that the priority of these additional services is very low. Figure 4.48: workspace layout when stationary nodes are introduced

35 Page 34 Figure 4.48: stationary additional nodes throughput Figure 4.49: stationary additional nodes delay

36 Page 35 Figure 4.50: moving low demand nodes throughput 5. Supplemental Analysis Figure 4.51: moving low demand nodes delay To answer the third question we posed in our Simulation model we must compare the quality of service achieved on our chosen n standard as compared to other standards at varying data rates. For this purpose we collect global throughput and delay statistics in our sample space model. From the global delay perspective we observe that the max delay for our basic sample space model is attributed to the b wireless standard at a data rate of 18 Mbps.

37 Page 36 Below that the max delay is attributed to two data rates of the g standard where the data rates in question are set at 18Mbps and 24Mbps. However the least delay is observed by the n 2.4 GHz standard from the list of chosen standards we analyses. What is interesting to note here is that although g has a case where the data rate set at 24Mbps is higher than n at 19Mbps we see that there is significant difference in delay in favor of the n standard. Figure 5.1: global delay When we analyze the Global throughput for the same we observe the max throughput to belong to the n standard where the higher the data rate has a higher throughput. Further it can be observed that the b standard has the lowest throughput from all the standards and to make a case in point we have run the b standard at its maximum data rate of 11Mbps.

38 Page 37 Figure 5.2: global throughput 6. Results Here we seek to answer the questions we posed in the Riverbed simulation model. In order to do so we acknowledge that on the basis of our basic simulation model concerning the 4 primary stationary nodes the highest throughput gained was for our skype user followed by the YouTube video streamer, the counter strike game player and then the web browser. In the terms of delay we notice that the descending order of delay is the YouTube user followed by the skype user, the gamer and then the lowest delay was for the web browser. We attribute the max delay to belong to the YouTube user due to complex nature of high resolution videos and the fact that we set it to 30fps. From our list of selected new users we expect the low resolution video user to have the highest throughput followed by the gamer, the VoIP user and then the users introduced in our final case 4. When examining the individual graphs for 4 primary stationary nodes we notice that on each occasion the application with the highest throughput i.e the skype application which has at minimum 4 times the throughput of others decrease the most. What is essential to notice here is that on comparing the throughput of the skype application we see that the decrease is the most when we introduce the low resolution video followed by the gamer and then the VoIp application. However the low throughput Http application had next to none effect.

39 Page 38 This observation is in line with our estimation of the effect of the new users based on their throughput. Further we observed that based on the type of service or priority level of the incoming user there was a corresponding fall in throughput and increase in the delay of the primary user using the same type of service. For example we notice that when the new video user is introduced the YouTube user experiences in throughput and increase in delay. The same can be observed in the gaming and VoIP case. The counter effect is that those applications which are high throughput and not on the same type of service as the new user experience an increase in throughput and decrease in delay. For example in our case 2 where the new gamer is introduced the throughput for the existing primary gamer goes down with significant increase in delay. A perfect example is for our case 3 where both the gamer and YouTube user experience an increase in throughput and decrease in delay when the VoIP user is introduced with the same type of service and priority as the Skype user. The cause for this may lie in the introduction of MIMO technology in the n standard that allows transmitting data via multiple antennas. Continuing on as the new user moves around, our sample space we notice a trend of there being an increase in the delay of each of our 4 primary application including the low throughput Http cases. At the same time the new user experienced much more spiking and inconsistencies in the throughput of the application. As for our choice of standard, the supplemental analysis shows that n at the chosen frequency and data rate is a good choice for our analysis. 7. Future Work For Future work we would foremost like to include various new standards especially the ac standard. Furthermore we would like to replace the pre-defined applications on Riverbed modeler with custom applications designed by packet capturing the relevant data. A third thing we would like to explore is to have more access points and have significantly more traffic flowing through them to examine the real life scenario of a campus as large as SFU. 8. References [1] Vangie.B, What is wifi, [Online] (Accessed Feb 12, 2015) Available: [2] National Instruments "WLAN A,b,g and N." [Online] (Accessed Mar 12, 2015) Avaliable:

40 Page 39 [3] Microsoft, What is QoS, [Online] (Accessed Feb 12, 2015) Available: [4] Topher.K Diagnosing and addressing Wi-Fi signal quality problems, [Online] (Accessed Feb 12, 2015) Available: [5] Learning Centre Wireless Networking, [Online] (Accessed Feb 12, 2015) Available: [6] Khaled. A, Ljiljana.T Performance Analysis of VoIP Codes over Wi-Fi and WiMAX Networks [Online] (Accessed Mar 12, 2015) Available: [7] Kritika.S, Nitin.B, Namarta.K Performance Evaluation of WLAN Scenario in OPNET Modeler [Online] (Accessed Mar 12, 2015) Available: [8] Introduction to use OPNET Modeler [Online] (Accessed Mar 12, 2015) Available: [9] Rajesh.G, Srikanth.K WiFi Traffic Analysis Project Report [Online] (Accessed Mar 12, 2015) Available: [10] Wireless LAN Capture [Online] (Accessed Mar 12, 2015) Available: [11] Sophia.C, Curtis.R, Thomas.S Performance Analysis of a Wireless Home Network [Online] (Accessed Mar 12, 2015) Available: [12] Kelvin.H, Titus.C, Glen.N Evaluation of Gaming Traffic Over WiMAX [Online] (Accessed Mar 12, 2015) Available: [13] Wireless Standards [Online] (Accessed Mar 12, 2015) Available: [14] Pi Huang Understanding IEEE ac VHT Wireless [Online] (Accessed Mar 12, 2015) Available: [15] Johannes Farber, Network game traffic modeling, Proceedings of the 1st workshop on Network and system support for games, p.53-57, Apr 16-17, 2002, Braunschweig, Germany