System Level Modeling of ieee 802. 16e Mobile Wimax networks: Key Issues




Дата канвертавання28.04.2016
Памер130.74 Kb.


System Level Modeling of IEEE 802.16e Mobile WiMAX Networks: Key Issues1
Raj Jain, Chakchai So-In, and Abdel-Karim Tamimi

Department of Computer Science and Engineering

Washington University in St. Louis
Saint Louis, MO 63130 USA

{jain, cs5, and aa7} @cse.wustl.edu

ABSTRACT

WiMAXTM has attracted a lot of attention recently in the telecommunication community including researchers, product developers and service providers. Numerous papers have been published on various design issues of WiMAX networks. Since the issues being discussed are usually ahead of the product availability, most of these studies require simulation. It is important to have some common features among these models so that their results can be compared. The Application Working Group at WiMAX Forum, therefore, has developed a standard simulation methodology that describes the key features to be simulated, the method of simulating these features and various parameter values to be used. This system level methodology has been used in several public and commercial WiMAX simulation models. This paper presents a summary of the simulation methodology and discusses the key issues and common mistakes in simulating various features of network configuration, physical layer, media access control (MAC) layer, and application layer.


1. Introduction

WiMAX Forum is a consortium of over 400 organizations interested in IEEE 802.16 based broadband wireless access. The members include service providers, equipment vendors, chip vendors, researchers, and users. The Forum develops interoperability specifications for equipment using IEEE 802.16 standards. IEEE standards have many options and allow a wide range of parameters. While this generality is good, it also makes it difficult for equipment from two vendors to interoperate unless they both choose the same set of optional features and similar values of various parameters. WiMAX Forum members limit the standard options to a set of profiles that can be implemented in products. The products are then tested for interoperability in WiMAX forum certification laboratories. WiMAX Forum certification ensures that the equipment purchased from two different vendors will interoperate. All networking and telecommunications technologies have similar interoperability organizations. For example, WiFi Alliance for IEEE 802.11 wireless local area network (LAN) standards.


It is important to understand the difference between IEEE 802.16 standards and WiMAX, in the sense that WiMAX networks implement only a subset of the features and parameter values allowed by the IEEE standard. For example, IEEE 802.16 allows OFDMA (Orthogonal Frequency Division Multiple Access) frame duration to be 0.5 ms, 1 ms, 2 ms, and more. WiMAX Forum has selected just one value: 5 ms. IEEE 802.16 allows mesh mode and license exempt mode operations. WiMAX Forum does not yet have a profile for these modes and so these modes are more of theoretical interest at this point in time. OFDMA frame duration has a significant impact on the throughput and delay performance of applications and so researchers analyzing WiMAX networks may get very different results just by using an incorrect value.
For performance results from two research studies or two vendors to be comparable, it is important that both be based on similar set of features and parameter values and that these choices be representative of the real-world equipment, which in the case of IEEE 802.16 equipment means parameters and features specified in WiMAX Forum profiles. In addition, new technologies often have features that are not well understood and may be modeled incorrectly. Therefore, experts need to discuss various modeling alternatives and select the best if there is one.
The Application Working Group at WiMAX forum is chartered to study the performance of various applications on WiMAX networks and to recommend best practices for optimizing the application performance. The tasks include developing standard application workload models. For the last three years, the working group has been developing a system level simulation methodology [2] that can be used by anyone interested in simulating WiMAX networks. The document is a result of three years deliberations by numerous simulation and technology experts. This paper presents a summary of the issues presented in that document.
Other wireless standard groups, such as, 3rd Generation Partnership Project (3GPP) and 3rd Generation Partnership Project II (3GPP2) also have similar simulation methodology documents [3-5]. IEEE 802.16m, the group developing the next generation of IEEE 802.16 networks, is also beginning its work with a standard modeling methodology [6]. This will facilitate comparison of various proposed alternatives to be based on a common set of assumptions.
The simulation methodology described in the WiMAX Forum document and summarized here is general so that it can be used with any modeling platform such as Network Simulator 2 (NS2) (www.isi.edu/nsname/ns ), CoWare (www.coware.com), OPNET (www.opnet.com), OMNET++ (www.omnetpp.org ), and Qualnet (www.qualnet.com).
The organization of this paper is as follows. Section 2 provides an overview of a Mobile WiMAX system and its key technology (OFDMA). Section 3 highlights the “system level” approach and “link level” approaches of modeling. Section 3 also provides a list of recommended values for important network configuration parameters. Section 4 addresses PHY modem abstraction for system simulation. Section 5 covers the MAC layer modeling issues and discusses scheduler design and modeling, which is a key to the performance of a WiMAX system. Section 6 captures the issues in application traffic modeling.
2. Overview of Mobile WiMAX and OFDMA
IEEE 802.16 supports a variety of physical (PHY) layers. Each of them has its own distinct characteristics. First, WirelessMAN-SC (Single Carrier) PHY is designed for 10 to 60 GHz spectrum. While IEEE has standardized this PHY, there are not many products implementing it because this PHY requires line of sight (LoS) communication. Also, rain attenuation and multipath affect reliability of the network at these frequencies. To allow non-line of sight (NLoS) communication, IEEE 802.16 working group designed the Orthogonal Frequency Division Multiplexing (OFDM) PHY using spectrum below 11 GHz [1]. This PHY is designed for fixed subscriber stations. WiMAX Forum has approved several profiles using this PHY. Most of the current WiMAX products implement this PHY. In this PHY, multiple subscribers use a Time Division Multiple Access (TDMA) to share the media. OFDM is a multi-carrier transmission in which thousands of subcarriers are transmitted and each user is given complete control of all subcarriers. For mobile users, it is better to reduce the number of subcarriers so that the subscriber station can use higher transmission power per subcarrier and increase their signal to noise ratio. This combination of time division and frequency division multiple access in conjunction with OFDM is called Orthogonal Frequency Division Multiple Access (OFDMA). The details of these interfaces can be found in [1].
IEEE 802.16 standard is designed to use any available spectrum width, which can range from 1.25 to 20MHz. Rather than designing a new circuit for each spectrum width, considerable cost savings can be obtained if the subcarrier spacing is kept fixed for all systems and the number of subcarriers is set proportional to the width of the available spectrum. This is called Scalable OFDMA or SOFDMA. Unless specified otherwise, 10 MHz spectrum width, a 5ms OFDMA frame time is used. The 5 ms frame time gives low latency jitter to delay sensitive applications while keeping the framing overhead also low. The default downlink to uplink subframe time ratio is 2:1. This is a compromise between data traffic which is highly asymmetric and voice traffic which is symmetric. For a voice-only network, downlink:uplink ratio of 1:1 will support more users. For data traffic, higher values of this ratio support more users.
3. System Level vs. Link Level Model
A link level model is used to study the transmission between a base station (BS) and one or more users. The key is to study the point-to-point link and the effect of various link level design decisions on the performance. The link level models generally concentrate on the PHY (physical) layer. Higher layers, such as IP routing, TCP sessions, or applications may or may not be modeled. The performance is measured in terms of bit per second (bps) throughput. A system level model, on the other hand, consists of a network of base stations and the emphasis is on the application layer performance as expressed by user perceived quality of service parameters such as voice or video quality. The PHY layer is abstracted as much as possible. Figure 1 shows important issues in various layers. A link-level model concentrates on the PHY layer while the system level model concentrates on the other layers.


Figure 1: Component layers of a Model
3.1 Center Cell Configuration
It is well known that the performance of a wireless transmission is determined by the Signal to Noise Ratio (SNR). The noise usually arises from stray transmissions and thermal causes. However, in WiMAX as well as in other cellular networks, “interference” from nearby base stations may affect the signal much more than the noise and, therefore, it is important to model the Signal to Interference and Noise Ratio (SINR). To determine the interference, we need an accurate representation of various sectors of a cell and the surrounding cells. System level models, traditionally, use a center cell configuration consisting of 19 cells as shown in Figure 2.


Basically the network consists of 19 hexagonal cells with six cells surrounding the center cell in the first tier. There are 12 cells surrounding the center cell in the second tier. Each cell has 3 sectors by default and thus there are 57 total sectors. A number of mobile subscribers (SS) are randomly dropped over these 57 sectors. Each subscriber station corresponds to an active user session that runs for the duration of a drop. A drop is defined as a simulation run for a given set of subscribers over a specified number of OFDMA frame times. At the beginning of each drop, the subscribers are associated with a specific base station and sector, henceforth referred to as the serving base station. The association is based on both the path loss and shadow fading, which are fixed for the duration of the drop.




Figure 2: 19-Cell Configuration used in System level Models



3.2 Interference Modeling
One of the key benefits of the OFDMA air interface is its possibility in enabling frequency reuse of one, i.e., the same frequency is used in all neighboring cells and sectors. This eases deployment since no frequency planning is required. This attribute is already available in CDMA networks, and is highly desirable in Mobile WiMAX based on IEEE 802.16e. High frequency reuse patterns, however, cause the system to become interference limited. The interference seen by a subscriber station in downlink and a base station in uplink is typically frequency selective and time selective. Therefore, it is not accurate to model interference as a white noise process with flat spectrum. The network simulators should model interference using a realistic channel model, which includes both slow fading and fast time-frequency selective fading components. As an optional simplification, the time-frequency channel of only the strongest interferers may be modeled to reduce simulation complexity. The remaining interferers are then modeled as (spatially) white and non-frequency selective Additive White Gaussian Noise (AWGN) processes whose variances are changing in time based on a Raleigh fading process.
3.3 Frequency Reuse Model
The network topology is basically divided in to clusters of N cells. Each cell in the cluster has a different frequency allocations; S sectors per cell and K different frequency allocations per cell. Thus, the frequency reuse pattern can be represented as N S K. Figure 3 shows a network topology with reuse pattern 1 3 3. The colored markings in the center of each cell indicate the sectors and point in the boresight direction. The red markings correspond to sectors deployment in the same frequency allocation. The blue and green markings indicate the other two frequency allocations for a reuse three network. Networks with universal frequency reuse 1 3 1 have the same network topology except the same frequency allocation is deployed in all sectors throughout the network. Thus, an operator using, say, 10 MHz channelization would require a total of 10 MHz of spectrum to support a time division duplexing (TDD) system with 1 3 1 reuse. To reduce interference, a frequency reuse pattern of 1 3 3 can be implemented by either sharing the available subchannels (say 1/3) in a 10 MHz channel or using 30 MHz of spectrum with 10 MHz in each sector.



Figure 3: Network topology for tri-sector (19 cell) configuration with 133 reuse
Table 1 shows a key subset of network configuration parameters that have been discussed in this paper so far. There are three sources for these recommended values. First, the values specified in the profile documents were used to set ranges. Second, the values used in recent requests for proposals by leading service providers were discussed to be used as default. Third, the members decided that the parameter ranges should be such that the models can be easily adopted for upcoming waves and the next generation of WiMAX based on IEEE 802.16m. This third requirement leads us to not limit the options to strictly follow the current profiles but be a superset of the current WiMAX profile. The default values are shown in bold and are compliant with current WiMAX profiles.

Table 1: Network Configuration Parameters

Parameter

Description

Value Range

Nc

Number of cells

19

S

Number of sectors/cell

1, 3, 4, 6

Ns = S×Nc

Total number of sectors

19, 57, 76, 114

R

Base station to base station distance

0.5 to 30km (1 km)

S

Number of sectors per Cell

3, 6

K

Number of frequency allocations in the network.

1, 2, 3, 4, 6

FBS

Frequency allocation (integer index) used in each base station sector

1, 2, 3, 4, 5, 6




Operating Frequency

2.0 – 3.5 GHz (2.5 GHz)




Duplexing Scheme

TDD


4. Physical Layer Modeling
In a system simulation, it is important to device a model that accurately predicts the link-layer performance in a computationally simple way. The requirement for such a model stems from the fact that simulating the physical layer links between multiples base stations and subscriber stations in a network/system simulator can be computationally prohibitive. This model is referred to as the PHY abstraction. The role of abstraction is to predict the decoded PDU Error Rate (PER) given a set of SINR values for each subcarrier k in the PDU and the Modulation Coding Scheme (MCS).
4.1 Equivalent SINR Computation
WiMAX uses OFDMA which is a multicarrier modulation. Each subcarrier has a different SINR). We need a method to combine these SINRs to a single equivalent SINR for an AWGN channel. Many wireless technologies use a simple method based on averages of SINRs. This is not adequate for an OFDMA system. This is because of four reasons. First, FEC block bits are spread between subcarriers. Second, due to frequency selectivity of desired signal, each subcarrier observes different SINR. Third, decoder behavior depends on the SINR fluctuations between FEC block bits, and not only on the average SINR. Fourth, bursts that observe different channel and interference characteristics will display different Bit Error Rate (BER) or Block Error Rate (BLER) results even though they may have the same average SINR. For OFDMA systems, the mapping from a vector of the code-word’s per-subcarrier SINR values to an AWGN-equivalent SINR metric is required
Several mapping functions have been proposed over the years: Quasi-static method, Convex method, Shannon method, and Exponential Effective SINR Mapping (EESM). After significant discussions, WiMAX Forum AWG group has decided to recommend EESM as a default recommended method. EESM mapping is defined as follows:

Where, SINRi is the SINR of the ith subcarrier in the code-word, β is an adjustment factor that depends on the FEC type, MCS, and modem implementation specific. WiMAX Forum commissioned a simulation study to determine the appropriate values of β. The recommended values are shown in Table 2. These values are to be used for SISO, PUSC mode, spatial channel model with the velocity of 3 m/h (Ped-B) and 60 km/h (Veh-A) and 100 independent channel realizations. Ideal channel estimation is assumed.


Table 2: Reference EESM Beta Values

Index

Mod-
ulation


M
= Bits
per
symbol


Rate

n
=Repetitions


J

Data
block
size


(bytes)

Encoded

data block

size (bytes)
=48×n×M/8



Ped-B
(dB)



Veh-A
(dB)


1

QPSK

2

1/2

1

10

6

12

2.46

2.54

2

QPSK

2

1/2

2

10

12

24

2.28

2.26

3

QPSK

2

1/2

3

10

18

36

2.27

2.26

4

QPSK

2

1/2

4

10

24

48

2.18

2.12

5

QPSK

2

1/2

5

10

30

60

2.05

2.07

6

QPSK

2

1/2

6

10

36

72

2.00

2.06

7

QPSK

2

1/2

8

10

48

96

2.03

2.02

8

QPSK

2

1/2

9

10

54

108

2.04

2.01

9

QPSK

2

1/2

10

10

60

120

1.98

2.01

10

QPSK

2

3/4

1

6

9

12

2.56

2.50

11

QPSK

2

3/4

2

6

18

24

2.43

2.43

12

QPSK

2

3/4

3

6

27

36

2.46

2.44

13

QPSK

2

3/4

4

6

36

48

2.41

2.39

14

QPSK

2

3/4

5

6

45

60

2.41

2.41

15

QPSK

2

3/4

6

6

54

72

2.38

2.37

16

16-QAM

4

1/2

1

5

12

24

7.45

7.48

17

16-QAM

4

1/2

2

5

24

48

7.14

7.14

18

16-QAM

4

1/2

3

5

36

72

7.00

6.92

19

16-QAM

4

1/2

4

5

48

96

7.34

7.53

20

16-QAM

4

1/2

5

5

60

120

6.89

6.82

21

16-QAM

4

3/4

1

3

18

24

8.93

8.93

22

16-QAM

4

3/4

2

3

36

48

8.87

8.87

23

16-QAM

4

3/4

3

3

54

72

8.85

8.90

24

64-QAM

6

1/2

1

3

18

36

11.31

11.43

25

64-QAM

6

1/2

2

3

36

72

11.11

11.16

26

64-QAM

6

1/2

3

3

54

108

11.09

11.01

27

64-QAM

6

2/3

1

2

24

36

13.80

13.74

28

64-QAM

6

2/3

2

2

48

72

13.69

13.70

29

64-QAM

6

3/4

1

2

27

36

14.71

14.68

30

64-QAM

6

3/4

2

2

54

72

14.59

14.55

31

64-QAM

6

5/6

1

2

30

36

15.32

15.17

32

64-QAM

6

5/6

2

2

60

72

15.29

15.27


5. MAC Layer Modeling
The main functions at MAC layer that have significant impact on the performance of WiMAX systems are: Scheduling, Automatic Retransmission Request (ARQ) and Hybrid ARQ (HARQ), fragmentation and packing, and Packet Header Suppression (PHS). Among these, scheduler is possibly the most significant and complex. If a wrong scheduler is used, it is possible to get meaningless results from simulation models. Also, IEEE 802.16 standard does not specify any scheduling mechanism. In this section, we discuss the simulation issues related to scheduling and refer the reader to the WiMAX simulation methodology document for discussion of other MAC features.
The MAC scheduler must efficiently allocate the available radio resources in response to bursty data traffic, time-varying channel conditions, and specified scheduler criteria, if any. The data packets are associated to service flows with well defined QoS parameters in the MAC layer so that the scheduler can correctly determine the packet transmission ordering over the air interface. The CQICH provides fast channel information feedback to enable the scheduler to choose the appropriate coding and modulation for each allocation. The adaptive modulation and coding combined with ARQ and HARQ provide robust transmission over the time varying channel.
The scheduling service is provided for both downlink and uplink traffic. In order for the MAC scheduler to make an efficient resource allocation and provide the desired QoS in the uplink, the subscriber station must feedback accurate and timely information as to the traffic conditions and QoS requirements.
The scheduler basically first chooses the appropriate coding and modulation for each allocation and then maps individual service flows to specific regions of the 2-D TDD frame. The mapping problem is a generalized 2-dimensional bin packing problem (NP-hard) thus heuristic schemes are used for practical systems, typically vendor-specific and proprietary implementation.







Figure 4: Downlink Packet Scheduler (left) and Uplink Packet Scheduler (right)
Downlink scheduler basically allocates the resource in each frame based on the traffic conditions and QoS requirements of service flows at subscriber stations. As shown in Figure 4, the scheduler needs to maintain per-flow queues so that QoS of each flow can be met. Uplink scheduler is very similar to downlink scheduler. Uplink scheduler maintains the request/grant status of various uplink service flows. Bandwidth requests arriving from various uplink service flows at the WiMAX, base station will be granted in a similar fashion as the downlink traffic. One additional consideration is that of available subscriber station power. This power is divided among allocated subchannels and therefore, affects the number of subchannels that the subscriber station can use. Another important consideration is that the left over resources should be fairly allocated. A quantitative measure of fairness is defined in [9].
6. Application Layer Modeling

There are two main types of traffic models: foreground traffic and background traffic. The foreground traffic represents a specific user behavior or interaction. On the other hand, background traffic is not directly related to a specific user interaction or activity.


Each model can be represented using two major traffic models: user level traffic model and IP packet level traffic model. An example of user-level model is the distribution of video frame sizes. The performance metrics are also at the user level, e.g., the number of video frames with error. IP packet models represent the packet arrival statistics at the IP layer. These models can be generated from network traffic measurements.
User level model is more complicated than IP level model but gives more information about users’ Quality of Experience (QoE). WiMAX Forum has identified the following application classes to cover several applications for 802.16e-based systems: Multiplayer interactive gaming, voice over IP (VOIP) and Video conference, Streaming media, and web browsing and instant messaging. Three key issues in performance modeling are:

  1. The standard traffic models specified in older technologies such as 3GPP and 3GPP2 documents are now dated. Web usage model is a clear example. 3GPP documents have a model of web traffic [10]. However, with the advent of Web 2.0, the web usage pattern has changed significantly. The older models no longer correctly represent the current user behavior.

  2. No traffic models are available for many of the new applications, e.g., gaming, video streaming, instant messaging, and so on.

  3. Everyone uses a different model for the same application and comes up with very different conclusions.

To avoid such problems, WiMAX Forum has selected a set of traffic models that are recommended to be used in system level studies. These models include: Internet gaming, VOIP, video conferencing, push-to-talk, music/speech, small video clips, movie streaming, multimedia broadcast services, instant messaging, web browsing, email, telemetry, FTP (file transfer protocol), peer-to-peer networking, VPN (virtual private networking) service, and near real-time video. These models are based on recent measurements and represent current user behavior.


7. Conclusions
System level modeling of IEEE 802.16e based mobile WiMAX networks requires a careful consideration at every layer including physical layer, MAC layer, and application layer. It is possible to come up with misleading results unless careful attention is paid to various key features at these layers. Even network configuration and user placement are important. For this reason, WiMAX Forum has developed a system-level simulation methodology. Along with pointing out how to model some key features, the document also points default parameter values that should be used. Anyone involved in simulation modeling should use this methodology so that their results can be compared with those of others.
The summary guidelines presented in this paper can be used by WiMAX service providers, equipment vendors, and users in their performance evaluation projects. These guidelines are general in that it can be used for any kinds of simulation tools such as NS2, OPNET, OMNET, and Qualnet. In this paper, we discussed many of the WiMAX system modeling issues: physical layer modeling, MAC layer modeling, and application modeling. For more details on these and for advanced topics, such as, modeling of Multiple-Input Multiple-Output (MIMO), Adaptive Antenna System (AAS), and Beam Forming, the reader is urged to refer to the detailed WiMAX Forum document.
Acknowledgements
WiMAX System Level Methodology is a result of three years of work by many members of the Application Working Group. Contributors include Ali Koc (Intel), Arun Ghosh (AT&T), Arvind Raghavan (Arraycom), Honghai Zhang (Lucent), Hyunjeong hannah Lee (Intel), Jaeyoung Kim (Samsung), Jeff Andrews (University of Texas), John Kim (Sprint), Jungnam Yun (POSDATA), Krishna Kamal Sayana (Motorola), Krishna Ramadas (Venturi Wireless), Maruti Gupta (Intel), Louay Jalloul (Beceem), Maruti Gupta (Intel), Mineo Takai (Kozo Keikaku Engineering, Inc.), Muthaiah Venkatachalam (Intel), Nat Natarajan (Motorola), Pete Gelbman (Clearwire), Raj Iyengar (RPI), Rok Preseren (Telsima), Roshni Srinivasan (Intel), Sherry Chen (Intel), Shyam Parekh (Lucent), Teck Hu (Siemens), Tom Tofigh (AT&T), Vafa Ghazi (CoWare), Xiangying Yang (Intel), Yaron Alpert (Comsys), and Zee'v Roth (Alvarion). Jie Hui (Intel) edited the MAC layer modeling section and Bong Ho Kim (POSTDATA) contributed most of the application models. The  values listed in Table 2 are from a study conducted by Bejing University of Post and Telecommunications (BUPT). We would like to specially thank Dr. Mostafa (Tom) Tofigh – Chair of the Application Working Group for giving us an opportunity to participate in this process.
8. References
[1] IEEE P802.16Rev2/D2, “DRAFT Standard for Local and Metropolitan Area Networks, Part 16: Air Interface for Broadband Wireless Access Systems,” December 2007, 2094 pp.
[2] WiMAX FORUM, “WiMAX System Evaluation Methodology V2.0,” December 2007, 230 pp.
[3] 3GPP, “Spatial Channel Model for Multiple Input Multiple Output (MIMO) Simulations,” Release 6, TR25.996, V6.1.0, September 2003.
[4] 3GPP2, "CDMA2000 Evaluation Methodology," V1.0, December 2004, 194 pp., available at www.3gpp2.org/Public_html/specs/C.R1002-0_v1.0_041221.pdf
[5] NGMN, "Next Generation Mobile Networks Radio Access Performance Evaluation Methodology," June 2007, Available at http://www.ngmn.org/docs/NGMN_Evaluation_Methodology_V1.2.pdf
[6] Roshini Srinivasan (Editor), “Draft IEEE 802.16m Evaluation Methodology”, December 2007, Available at http://www.ieee802.org/16/tgm/docs/80216m-07_037r2.pdf
[7] V. Erceg, K.V.S. Hari, et al, “Channel Models for Fixed Wireless Applications,” IEEE 802.16 Broadband Wireless Working Group, July 2001.
[8] ITU-R Recommendation M.1225, “Guidelines for Evaluation of Radio Transmission Technologies for IMT-2000,” 1997.
[9] Raj Jain, “The Art of Computer Systems Performance Analysis: Techniques for Experimental Design, Measurement, Simulation, and Modeling,” Wiley-Interscience, 1991.
[10] 3GPP2-C50-EVAL 2001 022-0, “HTTP and FTP Traffic Models for 1xEV-DV Simulations,” Lucent, Motorola, Qualcomm, Samsung contribution, 2001.

1 This work was sponsored by a grant from WiMAX Forum.



База данных защищена авторским правом ©shkola.of.by 2016
звярнуцца да адміністрацыі

    Галоўная старонка