How Quality of Service (QoS) is achieved in WiMAX (IEEE 802.16) Johan ElgeredA, 1 , A.Safaei MoghaddamA, 1 , and Benjamin VedderB, 1 A B

Department of Computer Science, Chalmers University of Technology Department of Signals and Systems, Chalmers University of Technology 1 {elgered, safaeia, vedder}@student.chalmers.se

Abstract—This report deals with Worldwide Interoperability for Microwave Access (WiMAX) technology with focus on Quality of Service (QoS). The basics of the technologies for the physical layer and the Media Access Control (MAC) layer are introduced. Also, a simulation was made to evaluate how the Best Effort (BE) scheduler performs in different scenarios. Index Terms—WiMAX, QoS, IEEE 802.16, MIMO, OFDMA, ns-2, Best Effort scheduling

I. I NTRODUCTION We are always in the need for higher communication speed. This need for more speed is not just in transmission speed of information, but also we like to move faster and do things faster. In other words be more efficient, which mobile communication offers. The need of mobility alongside with a high data transfer speed brings the essential need of a new technology in wireless communication to the picture. In addition, availability is another essential aspect in today’s communication. All this means that wireless communication is becoming more and more important. Worldwide Interoperability for Microwave Access (WiMAX) technology was founded in 2006 by Korea Telecom to fulfil these needs [1]. WiMAX is based on the 802.16 standard and overseen by Institute of Electrical and Electronics Engineers (IEEE). In the development of WiMAX various changes had to be done in different layers of the OSI (Open Systems Interconnection) model [8] in 3rd Generation Mobile Telecommunications (3G). A bigger area coverage with an increase in transmission speed was introduced in 802.16d also know as WiMAX base. WiMAX base introduced a dynamic bandwidth assignment to different mobile users, based on their needs at a specific moment. Though, it did not solve the problem of high rate data transfer to a moving mobile node with a vehicular speed. In 802.16e-2005 also known as Mobile WiMAX,

additional technologies were supported to guarantee a minimum Quality of Service (QoS) while the mobile is on move. This latest WiMAX version has pushed the boundaries of wireless communication further. Wireless communication has become a widely spread method to exchange information, leading to more interactive information and thus a larger requirement of bandwidth. An important factor in the competition between different wireless technologies is the capability of meeting Quality of Service (QoS). QoS is defined as the performance guarantees a network system can make regarding packet loss, delay, throughput and jitter [2]. Different frameworks have been developed within the field of QoS; Integrated Services (IntServ) provides individualized QoS guarantees to particular flows (packet stream with common source address, destination address and port number), while Differentiated Services (DiffServ) divides flows into separate classes [3]. Controversy exists whether QoS is needed or not. Those who are against it state that when traffic has reached a level beyond the capacity of the network, QoS will not manage to satisfy user demands and if the network has enough resources for all traffic, QoS is unnecessary. Those who argue for QoS are of the opinion that it makes the information exchange more fair by dividing resources among users and also that it allows the network to run with a more intensive usage. The main topic of this paper covers mostly 802.16e, where the focus lies on the lower layers connected with QoS. These layers include the physical layer and the Media Access Control (MAC) layer. A simulation was also made to analyse how the Best Effort (BE) scheduler performs. II. P HYSICAL LAYER The physical layer in mobile WiMAX has two key technologies; Orthogonal Frequency-Division Multiple

munication systems. The capacity of the SISO additive white Gaussian noise (AWGN) channel is [5]: C(t) = log2 (1 + ρ|H|2 )bits/s/Hz

(1)

Where H is the channel matrix and ρ is the average signal-to-noise ratio (SNR). For this case one extra bit for the capacity needs 3dB extra power. For MIMO, when assuming the transmitter has M antennas and the receiver has N antennas and Channel Side Information (CSI) is not present, the capacity is [5]: ρ HH ∗ )]) N ≈ αmin(M, N )bits/s/Hz

C = εH (log2 [det(IM +

Fig. 1: Illustration of differentiation between SISO, SIMO, MISO and MIMO wireless communication systems [4]

Access (OFDMA) and Multiple-input and Multipleoutput (MIMO). In the physical layer the delay, throughput and jitter regarding QoS can be affected. MIMO is the use of multiple antennas at both transmitter and receiver. For example, 4 x 2 MIMO means using four antennas at the base station and two antennas at the mobile device. An illustration about different antenna setups is shown in Figure 1. Comparing with the traditional Single-input Single-output (SISO) antenna model, MIMO increases the capacity and the spectral efficiency of the communication system greatly. Therefore, the system offers higher data rate with limited bandwidth, which increases the throughput regarding QoS. MIMO can deal with the multipath fading easily, however, it has no solution for frequency selective fading. OFDMA is a multi-user technology which is the evolution of Orthogonal Frequency-Division Multiplexing (OFDM). Mobile WiMAX uses MIMO together with OFDM to achieve good performance on frequency selective fading channels. Although OFDM has a lot of advantages compared to other traditional modulation schemes, there is no obvious improvement on the channel capacity. On the other hand, MIMO could increase the channel capacity by increasing the number of antennas. So the technology of MIMO combined with OFDM could offer a stable and low error service. Given these facts, the MIMO-OFDM model should be one of the most promising and widely used technology in the future. Capacity is a very important measurement of telecom-

(2)

where α is a constant, H is the channel matrix and ρ is the average SNR at each receiver branch. As the equation shows, the capacity is proportional to the minimum number of transmitter and receiver antennas. Therefore, it can be seen that when comparing to SISO, MIMO has great potential in increasing the capacity when the bandwidth and the power are limited, which is very valuable in telecommunication transmission [6]. A. Transmission models The 802.16 standard supports three MIMO models which are Matrix A (Transmit Diversity rate = 1), Matrix B (Transmit Diversity rate = 2) and Matrix C (Highest rate, times 4). Diversity gain can also be achieved by using Space Time Coding (STC) without transmitter CSI. With STC, a single data stream is transmitted over multiple antennas. Different data bits are transmitted over different antennas during the first symbol period and the conjugate or inverse of the same bits are transmitted again. This way, the received signal is more robust while the rate is unchanged. STC can be used to enhance the coverage area and as a better channel allows for higher order modulations even the capacity can be increased. Another method is to transmit different independent data streams over different transmitter antennas. This is the technology known as Spatial Multiplexing (SMX). If the system could separate different data streams well, it would behave just like parallel channels. Therefore the data rate is increased. The spatial multiplexing method is important as it can increase the capacity without additional power or bandwidth consumption [7]. Given multiple transmitter and receiver antennas, STC technology can be used combined with SMX technology.

How to combine STC and SMX is a trade off whether higher data rate or lower bit error rate is preferred. So, the combination of STC and SMX affects the delay, throughput and jitter regarding QoS. B. IEEE 802.16e-2005 Mobile WiMAX attacks the problem of mobility in wireless communication. This brings the possibility of high data transfer speed at vehicular speed mobility. Also, it supports dynamic assignment of multiple modulations for both robust and high speed data transfer rate such as Quadrature Phase Shift Keying (QPSK) and Quadrature amplitude modulation (QAM). As the mobile device moves further away from the station, the modulation changes, providing a more robust connection with lower bandwidth. By contrast, as the mobile moves towards the station, more channels are assigned to that device in order to provide a higher bandwidth. 802.16e-2005 can use more sub carriers with OFDM and OFDMA with 128 sub carrier Fast Fourier Transform (FFT) up to 2K sub carrier FFT. Therefore, a higher speed rate can be offered [11]. III. MAC L AYER The MAC layer is a sub-layer to the Data Link Layer which exists in the layer 2 of the OSI model [8]. The MAC layer refers to the different protocols and mechanisms to allow simultaneous users to access a specific network, in this case a Mobile WiMAX network. It consists of three sub-layers: the convergence sub-layer (CS), the common part sub-layer (CPS) and the security sublayer [9] (see Figure 2). CS approves arriving Protocol Data Unit’s (PDU) from the higher layer, called MAC Service Data Unit’s (MSDU). It is within this layer that the fundamental QoS mechanism in 802.16 is found, in the shape of functions of classifying and mapping MSDU’s into adequate Connection Identifier’s (CID). After the CS has performed mapping and classification, it transports the MSDU’s to CPS where common MAC functions are executed to achieve QoS, for example fragmentation, packing and concatenation. A. IEEE 802.16m Because WiMAX is competing with LTE in the 4G area, a group has been developing 802.16e further. This additional system to the IEEE 802.16 standard is called 802.16m, also known as WiMAX 2. A positive element with WiMAX is that the system is backwards compatible why 802.16e and 802.16m works in parallel. Thus,

Fig. 2: Protocol layers of WiMAX [10]

mobile devices where some uses 802.16e while others use 802.16m can be served by the same base station on the same carrier. The difference between them is that 802.16e uses one single carrier while 802.16m can use several carriers to increase data transfer rates. The requirement of downlink data rates in 802.16m is 100 Mbps in mobile and 1 Gbps in stationary. WiMAX 802.16m was recently approved as a standard by IEEE and will be introduced to the market 2012 [12]. Its basic improvements compared to 802.16e are the support for new service classes, increased mobility and better Quality of Service gurantees [13]. B. QoS in 802.16m In the IEEE 802.16m QoS is achieved by the interpretation of flows of packets as service flows. Just like in the previous version, 802.16e, the flows occur in a single direction and each flow is mapped to one transport connection. To increase QoS, the new features in this standard are the improvements of the polling and granting mechanisms. Several sets of QoS parameters can be chosen for each service flow. This is achieved in the interplay between Mobile Station (MS) and Base Station (BS), which negotiate the QoS parameters when the service flow is configured. The process is executed in a flexible way because both BS and MS can separately adjust the QoS parameters dynamically, according to a predefined set if different QoS requirements would appear or if the traffic characteristics would change. An example of this is in the use of Voice over IP (VoIP) applications. Initially when the connection between the BS and MS is established, two different sets of QoS parameters can be decided. One set could be used during

periods when voice is transported over the connection and another set could be used during silent periods. Through these manners, QoS is achieved by managing the conditions that arise during the changes of states an application pass through [13]. C. QoS Scheduling

module [14]. The extended WiMAX module was developed to add support for further scheduling services more than just the existing scheduler included in the module, which works as a BE scheduler. A comparison was made between the original BE scheduler and the new implemented BE scheduler, in a setting with several rtPS Subscriber Station’s (SS) and two BE SS:s. The outcome of the simulation was that the original BE scheduler provided the same BE throughput independently of the traffic load. In comparison, the new implemented BE scheduler provided lower throughput as the traffic load was increased. The conclusion drawn of this was that the new implemented BE scheduler was more realistic since BE connections do not have the strict QoS requirements as other scheduling services. In another study, simulation was performed with the aim to compare the different scheduling services UGS, rtPS and ertPS [15]. The performance metrics used in the simulation were system throughput, packet delay and signaling overhead as a function of traffic load in the system. The simulation was done in Matlab and consisted of movement algorithms for the SS:s to never move beyond the coverage area of the BS:s. Also, different traffic models were used to make the comparison between the scheduling services fair (eg. ertPS performance strongly depends on the rate of traffic size changes). The results showed that UGS is not a good alternative because overall system capacity tended to decrease alot in its use. When ertPS was in use the best results was obtained; system throughput was considerable higher, whereas signaling overhead and packet delay were smaller in comparison to the other test cases.

There exists five different scheduling service types in 802.16 standard MAC layer. Through these, WiMAX offers the ability to give different applications varying size of bandwidth associated with the priority of the application, in comparison to WiFi where each application has the same priority. It is the BS scheduler that distributes the amount of bandwidth necessary for an application, depending on which QoS class the application has. With the QoS parameters (see Table I) given, the BS can offer polls and or grants at adequate occasions. The five scheduling service types are presented below. 1) UGS (Unsolicited Grant Service): When a UGS scheduling service type is used, the BS grants data packets, at periodic intervals, which are fixed-sized. Thus, it is suitable for real-time data streams such as VoIP. 2) rtPS (Real-Time Polling Service): Has the same function as UGS, but where UGS grants fixed-sized data packets, ertPS uses dynamic distribution. An example of an application is transportation of video, for example MPEG (Moving Pictures Experts Group). 3) ertPS (Extended Real-Time Polling Service): This scheduling service type is based on both UGS and rtPS. It is adequate for applications as VoIP, in which the data rate is variable. 4) nrtPS (Non-Real-Time Polling Service): Requires minimum data rates for data packets with variable size. It is made with the intention of managing delay and is therefore convenient in the use of File Transfer Protocol (FTP). The BS regularly offers request polls, which makes it possible for the application to make requests even though network congestion is in progress. 5) BE: This scheduling mechanisms can cause long delays when there is network congestion since it handles applications on best available basis. It has no support for applications which have requirements of minimum service guarantees. An example of an application that uses this scheduling service is E-mail. [10]

For the simulation the ns-2 simulator [16] was used along with the NIST WiMAX module [17]. The ns-2 simulator is principally targeted for network research. It has been under development since 1989 and is still in progress. Ns-2 uses two languages, C++ and OTcl. C++ is suitable for detailed implementation of protocols because it is fast to run. Though, changing settings using C++ would be rather slow why OTcl is to prefer, where changes can be made interactively. The developers state that it is not a polished and finished product and bugs are being discovered and corrected. However, many studies are based on the use of the ns-2 simulator.

IV. C OMPARISON OF SCHEDULING SERVICES

VI. S IMULATION MODEL

In one study, simulations of UGS, rtPS and BE were made with the ns-2 simulator and an extended WiMAX

Simulation was done regarding packet loss in wimax network in two scenarios. Packet loss is described as

V. W I MAX NS-2 M ODULE

Scheduling service UGS rtPS nrtPS BE

Maximum sustained traffic rate • • • •

Minimum reserved traffic rate (Can be present) • •

Request/transmission policy

Tolarated jitter

Maximum latency

• • • •

•

• •

Traffic policy

• •

TABLE I: QoS parameters of the 802.16 scheduling services [10]

the difference between the total packets transferred by the base station and the total number of packets received by all nodes. In order to evaluate the BE scheduler regarding mobility, two scenarios were chosen, scenario A and B. In both scenarios an omni antenna is used. The bit rate chosen was 11 Mb/s. Also, a 0.02 second gap was used between packet transmission of each node to avoid packet loss due to BW request collision. The base station coverage area was 500m radius. In scenario A, all nodes are in the same distance of the antenna in a static position. The simulation was executed with an increasing number of nodes from the beginning towards the end. The code below are examples of how some of the settings were set.

Fig. 3: Simulation with up to 200 static nodes.

# disable random motion $wl_node_($i) random-motion 0 ; $ns at 0 "$wl_node_($i) setdest 550.0 550.0 40.0" # network interface type set opt(netif) Phy/WirelessPhy/OFDM ; # MAC type set opt(mac) Mac/802_16/BS ; # antenna model set opt(ant) Antenna/OmniAntenna ; # Traffic scenario: if all the nodes start talking # at the same time, we may see packet loss due to # bandwidth request collision set diff 0.02 for {set i 0} {$i < $nb_mn} {incr i} { $ns at [expr $traffic_start+$i*$diff] "$cbr_($i) start" $ns at [expr $traffic_stop+$i*$diff] "$cbr_($i) stop" }

VII. S IMULATION RESULTS As Figure 3 shows, as the number of nodes increased, the packet loss would increase also. In scenario B all the parameters were the same in the beginning. Over time the nodes would start moving towards the base station with a speed of 40m/s and then stop at the base station. Figure 4 shows that the packet loss increased compared to figure 3. Packet loss may also occur due to collision. To avoid this a 0.02 second gap between packet transmission of each node was added, however, it may still occur. Thus, the packet loss can be described as the following:

Fig. 4: Simulation with up to 200 nodes moving towards the base station.

P L = Ptx − Prx − Pbwc

(3)

Where P L is the packet loss of interest, Ptx is the total packets transmitted, Prx is the total packets received and Pbwc is the packets lost due to bandwidth collision. Consider P L1 as the packet loss from scenario A and P L2 as the packet loss in scenario B. The difference

between PL1 and PL2 is a considerable number in percentage of the whole transferred packets. By comparing PL1 and PL2, it could be said that the packet loss has increased in a considerable number when the nodes are moving around. In theory, a significant effect on the packet loss on moving nodes is due to the change of channel the attributes and therefore, the modulation and the size of the FFT that is used for data transfer. That means this is related to the physical layer and not the scheduler. However this is true in reality, but there are reasons to take PL2 as a lack of the scheduler also. Firstly, in both scenarios the attributes of the physical channel is kept the same. Secondly, in scenario A despite that the nodes are static there is a packet loss, and this could be said to be related to the scheduler indeed. The authors reason on the increase in PL as to be related to the scheduler. In final, the simulation states that the BE scheduler does not satisfy the need of mobility with the given assumptions. Never the less, more reliable reasoning can be done when other schedulers have been also studied and simulated in a similar manner or scenario.

the physical layer, it also states that the BE scheduler lacks regarding mobility. In future work, the simulation should be done with more schedulers to confirm this. R EFERENCES [1]

[2]

[3]

[4] [5]

[6]

[7]

VIII. C ONCLUSION The paper shows how different aspects of QoS are met in mobile WiMAX in the physical layer and in the MAC layer. Also, the simulations made here support the statements of the paper. The physical layer has the, as of this writing, latest technologies using MIMO and OFDMA to achieve good performance regarding spectral efficiency, power consumption and most aspects of QoS. This means the physical layer in mobile WiMAX performs well and is up to date in comparison to other competing technologies. The MAC layer in Mobile WiMAX provides five different schedulers to optimize the use of different applications regarding QoS. The first scenario in the simulation shows that some packet loss occurs when many nodes are on the same base station and that some of this packet loss is not related to bandwidth restrictions. As no packet loss occurs at a lower number of nodes and the other parts of the scenario do not change with more nodes, this packet loss could most likely be changed with a different scheduler. However, to confirm this more schedulers should be evaluated with the same set-up. The second scenario in the simulation shows that even more packet loss, that is not related to the maximum bandwidth, occurs when the nodes are moving. The paper states that though this is most likely related to

[8] [9]

[10] [11] [12]

[13] [14]

[15]

[16] [17]

Mark Norris and Adrian Golds. Why WiMAX? URL: http://www.intercai.co.uk/library/pdf/Why- WiMAX. pdf. D. E. Comer. Internetworking with TCP/IP - Principles, Protocols and Architecture. 5th ed. Pearson Education, 2006. Worcester Polytechnic Institute, Department of Computer Science. URL: http : / / web . cs . wpi . edu / ∼rek / Adv Nets/Spring2002/IntServ DiffServ.pdf. Benjamin Baumgrtner. Prinzip MIMO. URL: http://en. wikipedia.org/wiki/File:Prinzip MIMO.svg. Sylvain Ranvier. Overview of MIMO systems. URL: http : / / www . comlab . hut . fi / opetus / 333 / 2004slides / topic47.pdf. Helmut Blcskei, David Gesbert, and Arogyaswami J.Paulraj. “On the Capacity of OFDM-Based Spatial Multiplexing Systems”. In: IEEE Transactions On Communications 50.2 (2002). Motorola. A Practical Guide to WiMAX Antennas: MIMO and Beamforming Technical Overview. URL : http : / / www . motorola . com / web / Business / Solutions/Industry%20Solutions/Service%20Providers/ Wireless%20Operators/Wireless%20Broadband/wi4% 20WiMAX/ Document/StaticFile/A Practical Guide to WiMAX Antennas White Paper.pdf. J. F. Kurose and K. W. Ross. Computer Networking - A Top-Down Approach. 4th ed. Pearson Education, 2008. Byeong Gi Lee and Sunghyun Choi. Broadband Wireless Access and Local Networks - WiMAX and WiFi. Artech House, 2008. Loutfi Nuaymi. WiMAX: Technology for Broadband Wireless Access. 1st ed. John Wiley and Sons, 2007. What is IEEE 802.16e? URL: http://www.wimax.com/ wimax-technologies-standards/what-is-ieee-80216e. Computer World. URL: http : / / www. computerworld . com/s/article/9215414/IEEE approves next WiMax standard?taxonomyId=15. et al. A. Bacciocola. “IEEE 802.16: History, Status and Future Trends”. In: Computer Communications (2009). Loutfi Nuaymi Aymen Belghith. Design and Implementation of a QoS-included WiMAX Module for NS-2 Simulator. Tech. rep. TELECOM Bretagne, 2008. Robert Bestak Pavel Mach. “Analysis and Performance Evaluation of IEEE 802.16 Enhanced with Decentrally Controlled Relays”. In: IWSSIP 2009. 2009. ns-2. URL: http://www.isi.edu/nsnam/ns/. Simulation Models for NS-2. URL: http://www.nist.gov/ itl/antd/emntg/ssm tools.cfm.

T HE REVIEW QUESTION Question How is Quality of Service (QoS) usually defined in network systems? Answer QoS is defined as the performance guarantees a network system can make regarding packet loss, delay, throughput and jitter.