BS coordination towards interference minimization in WiMAX networks

BS coordination towards interference minimization in WiMAX networks

Computer Networks 55 (2011) 356–369 Contents lists available at ScienceDirect Computer Networks journal homepage: www.elsevier.com/locate/comnet On...

1MB Sizes 0 Downloads 91 Views

Recommend Documents

Towards a common benchmark in WiMAX environment
Moves towards providing a common benchmarks in different network environments have recently been undertaken. Such suites

An interactive QoS framework for fixed WiMAX networks
The IEEE 802.16 standard defines several scheduling classes at MAC layer for preferential treatment of service flows dep

Sensitivity minimization in active RC networks
A general technique is presented to minimize sensitivity in the design of active RC networks. It is shown that the minim

On the impact of hybrid errors on mobile WiMAX networks
Burst and AWGN errors affect mobile WiMAX networks because of propagation conditions generally resulting from the mobili

Overlapping coalition formation games based interference coordination for D2D underlaying LTE-A networks
In the device-to-device (D2D) underlaying LTE-Advanced (LTE-A) systems, unlike the existing works which formulate the re

sGSA: A SDMA–OFDMA greedy scheduling algorithm for WiMAX networks
Future wireless networks need to address the predicted growth in mobile traffic volume, expected to have an explosive gr

Interference-aware broadcast scheduling in wireless networks
In this paper, we study the Interference -Aware Broadcast Scheduling problem, where all nodes in the Euclidean plane

Improved algorithms for latency minimization in wireless networks
In the interference scheduling problem, one is given a set of n communication requests described by source–destination

An efficient analytical model for the dimensioning of WiMAX networks supporting multi-profile best effort traffic
This paper tackles the challenging task of developing a simple and accurate analytical model for performance evaluation

Computer Networks 55 (2011) 356–369

Contents lists available at ScienceDirect

Computer Networks journal homepage: www.elsevier.com/locate/comnet

On BS/BS coordination towards interference minimization in WiMAX networks Christos P. Antonopoulos ⇑ INTRACOM S.A. Telecom Solutions, Panepistimiou 254, 26443 Patras, Greece

a r t i c l e

i n f o

Article history: Received 18 February 2010 Received in revised form 6 July 2010 Accepted 26 September 2010 Available online 1 October 2010 Responsible Editor: T. Tugcu Keywords: Broadband communications Communication systems Metropolitan area network Resource management Wireless MAN Network performance System network evaluation

a b s t r a c t Broadband wireless networking is a rapidly evolving research field supported by both industrial and academic communities. Focusing on the Medium Access Control (MAC) layer, IEEE 802.16 protocol family constitutes a prominent technology offering significant advantages in terms of meeting the continuously increasing application demands. However, cross-interference due to co-channel Base Station (BS) transmission residing in adjacent cells constitutes a challenging issue preventing optimum resource utilization in an IEEE 802.16 system thus significantly degrading network performance. In this paper a study on this problem is presented, based on a well-known system level network simulator. Furthermore, driven by valuable conclusions extracted by the aforementioned evaluation, an Internet Protocol (IP) layer coordination mechanism is presented, implemented and evaluated in the context of the network simulation environment, indicating significant behavioral and performance improvement. An important aspect of this work is to focus on critical concerns for industry, such as, the implementation feasibility and efficiency, while preserving backwards compatibility to existing systems. The main objective of the coordination mechanism is to achieve optimum bandwidth allocation in overlapping coverage areas, compensate from degraded throughput performance, and minimize side-effects due to coordination overhead. Ó 2010 Elsevier B.V. All rights reserved.

1. Introduction Over the last few years Wireless Broadband Access (WBA) systems, due to their appealing characteristics, have been receiving high interest by both academic and industrial societies. Such characteristics include, advanced MAC protocols (i.e. 802.16 protocol family [1]) as well as physical medium accessing techniques, i.e. Orthogonal Frequency Division Multiplexing/Access OFDM/OFDMA accessing techniques. Respective network systems claim to offer end-users enhanced network capabilities and performance in terms of available bandwidth, real time communication, quality of service (QoS), mobility etc. equal, or even superior, to xDSL services [2]. However, it was

⇑ Tel.: +30 2610 465033. E-mail address: [email protected] 1389-1286/$ - see front matter Ó 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.comnet.2010.09.012

apparent from early on that in order to harness, control and put these advanced characteristics in practical use, newly introduced techniques offering optimum resource utilization guarantees are of critical importance. Spatial division multiple access (SDMA) transmission techniques enable steering transmission/reception towards specific direction and covering areas of specific angular degree with beam-forming being a very well know and widely used method. This method is met in the context of cellular wireless systems as presented in [3] where joint power control and beam-forming schemes are proposed for cellular systems when adaptive arrays are used by the base stations. On the other hand [4] presents a research approach concerning the adaptation of SDMA transmission by an IEEE 802.16 based system along with the expected benefits from a respective SDMA scheduler, without however, considering the network reaction to co-channel interference by neighboring BSs or Mobile Stations (MSs) transmissions.

C.P. Antonopoulos / Computer Networks 55 (2011) 356–369

In the same paper an introduction is provided for 802.16 MAC protocol and SDMA through beam-forming antenna systems. Furthermore, high research interest is attracted by optimum scheduling approaches regarding a wide range of BWA network resources [5–11]. Following another approach, focusing mostly on PHY related issues, [5] considers the resource allocation problem of assigning set of sub-carriers and determining the number of bytes to be transmitted for each sub-carrier in OFDMA systems. Issues of interest in this work include simplicity, fairness and efficiency of the proposed algorithm. Also focusing on physical medium characteristics management but from a different perspective, [6] proposes adaptive transmission algorithms by exploiting the channel variation in space, and frequency domains. The evaluation is based on numerical results and complexity issues emerging in order to achieve enhanced performance. These efforts, and many more being available, present great interest and have offered significantly enhanced performance. However its becoming apparent that respective proposals impose issues of high complexity and implementation difficulty leading to significant reduction of the approach’s practicality and therefore its effective usefulness. As a consequence, parallel to the aforementioned efforts another perspective is gaining increasing interest. According to this view, resource scheduling is attempted at a higher network layer focusing primarily on the system wide performance. Based on this perspective, emerging proposals offer significant performance enhancement through practical, low complexity and adaptive methods/ techniques while retaining backward compatibility with already standardized protocols. Such a research effort is presented in [7] were the problem of packet transmission scheduling for real-time CBR traffic in IEEE 802.16 based mesh network is formulated and solved as a binary linear problem. Furthermore, aiming to lower complexity, and thus increase its practicality a heuristic suboptimal version is provided. The evaluation is based on numerical studies. On the other hand, efforts aiming toward interference management are of great importance and interest approaches. In this context [8] provides a very interesting, and shear commonalities concerning the objectives with the work presented in this paper. However, work presented in [8] although offering valuable input towards achieving interference minimization follows a different path in various critical considerations. Some of them concern hexagonal cellular structure and Subscriber Station (SS) able to electronically steer their beams towards the direction of the serving BS (which is a characteristic most probably not provided by actual devices) as well as assuming fixed SSs. Furthermore a distributed approach is presented as opposed to a centralized technique, presented in this paper, which the author considered a more feasible and efficient solution. Additional information would be useful concerning evaluation procedure as well as implementation details. The issue of interference minimization through packet scheduling is also addressed in [9–11]. The main objective of this paper is the proposal and respective evaluation of a non-hard real-time coordination mechanism able to efficiently arbitrate transmissions towards BSs that suffer co-channel interference. Non-hard

357

real time refers to the fact that coordination does not occur in strict per frame bases but rather respective period is in the order of a few frames (typically between 3 and 10 as will be indicated in the performance evaluation sections). Under this approach the main resource to be allocated is transmission time assuming that BSs are SDMA capable, so as to leave unaffected non-interference creating communications occurring concurrently. Effectively the problem in hand is caused by overlapping communication areas formed by adjacent BSs transmitting at the same frequency. Thus in the context of this research effort the final goal is to efficiently coordinate communications regarding overlapping areas in order to mitigate the unavoidable interference problem. Therefore in order to meet the aforementioned objective, firstly the significance of the problem must be evaluated. This step is critical since it enables the extraction of useful conclusions regarding behavioral characteristics leading to performance degradation. Such conclusions in essence will indicate, potential performance benefits from a transmission coordination approach as well as directions such an approach should follow. This evaluation is conducted based on the NS-2 WiMAX model [12] through substantial number of simulation scenarios varying critical network parameters so as to expose all possible and worth investigating behavioral cases. Secondly, the main conceptual approach will be presented along with detailed flow charts. Furthermore, concerning the implementation of this approach, the use of elements and entities already indicated in relative standards is considered, in order to advocate this approach as a realistic, feasible and efficient solution. Finally an implementation of the respective algorithm in the context of the NS-2 WiMAX model is tested considering the same scenarios as in the evaluation phase so as to provide objective comparison regarding the observed performance enhancement from this approach as well as the potential limitations commonly encountered in coordination efforts. As can be extracted by this introduction as well as the following sections the use of a network simulation platform and a WiMAX model is a critical one. Choosing to follow this path rather than, i.e. developing an analytical model was dictated by the main goal of this work, the complexity of communication protocol in hand, the complexity of the scenarios under evaluation and of course the validity of the WiMAX models existing nowadays. Therefore being the main goal to study and tackle a particular problem at system level makes the use of a simulation model the optimum solution so as to take into consideration all possible behavioral parameters of the communication protocol. Furthermore WiMAX protocol itself comprises by a high number of complicated functionalities and operations making the developing of respective analytical models at system level a task of high complexity without analogous benefits in terms of practicality and validity. At the same time the considered scenarios increase the complexity requiring multiple stations interacting, signal propagation model, specific physical transmission technique, wide range of transmission ratse etc. dictating even more the use of simulation model. Last but not least both the NS-2 system level simulator [21] and NIST WiMAX model [12] comprise well know and widely utilized simulation tools

358

C.P. Antonopoulos / Computer Networks 55 (2011) 356–369

able to accurately simulate complex WiMAX networks while providing the user to configuration a wide range of parameters. In particular the NIST model is based on the IEEE 80216e providing all required MAC and PHY functionalities for this research effort. Such functionalities include IEEE 802.16 framing and data transmission, OFDM physical layer with configurable modulation, cooperation with physical signal propagation model etc. This paper is organized as follows: in Section 2 a detailed performance and behavioral evaluation is presented regarding adjacent co-channel BSs, transmitting over common coverage areas. Through this evaluation the importance of the issue in hand is indicated while the design of coordination mechanism is driven. The design and specification of such an algorithm is provided in Section 3 emphasizing on implementation and feasibility issues. A simulation based evaluation of the proposed approach is presented in Section 4 regarding scenarios in correspondence to Section 2 scenarios in order to extract objective and useful conclusions. Finally the main conclusions and possible future extensions are summarized in Section 5. 2. The effect of co-channel interference The goal of this section is to present the effect of cochannel interference on network performance. This is attempted through sufficient number of simulation scenarios considering appropriate topologies where stations are affected by co-channel interference. Respective evaluation on network performance is based on measurements concerning throughput capability and mean packet delay being two cornerstone system metrics. Hence, adequate simulation scenarios include Mobile Stations (MSs) communicating with different BSs at the same frequency while residing in the overlapping area. The evaluation focuses on three adjacent, co-channel BSs deployment scenario forming the overlapping coverage area considering the following common parameters and performance metrics.  Constant parameters  Two ray ground propagation model considered 2 2 according to equation P r ðdÞ ¼ Pt Gt Gd4r ht hr where Gt, Gr are the antenna gains of transmitter and receiver respectively, Pt refers to the transmitter signal power, ht, hr denote the antenna heights of transmitter, receiver respectively while d indicates the distance.  Receive Threshold, 500 m radius.  OFDM transmission technique considered.  Same communication frequency in order to create co-channel interference conditions.  Large data packets (1500 bytes) with packet segmentation and reassembly enabled providing a more realistic user case.  Evenly distributed MSs to available BSs in order to equally distribute the aggregate workload.  No handover sequence is allowed so as to always be aware of traffic load allocated to each BS.  No Hybrid Automatic Repeat Request (HARQ) algorithm is considered that can alter system behavior.  Uplink direction.  System Network Simulator: NS-2 v 2.31.

 Variable parameters  Constant bit rate (CBR) traffic generators with packet creation intervals: 1–0.001 s per data flow corresponding to 12–1200 Kbps per flow. This range is selected for presentation since it presents a traffic rate ranging from relaxed scenarios up to scenarios clearly indicating saturation behavior therefore all the useful traffic cases.  Considered nNumber of MSs: 2–21 thus considering a wide range of possibly interacting stations enhancing the validity of the evaluation.  Main performance evaluation metrics Successfully Received at Application Layer ðBitsÞ  Throughput ¼ DataTime : Duration of Traffic Generation ðSecondsÞ  PacketDelay = TReceived  TCreated, where TReceived refers to the time the packet was successfully received at application layer and TCreated to the time the packet was send by application layer of the transmitter station.

Throughput graphs presented depict the ratio that mean throughout represent with respect to offered traffic. In this way not only the absolute measurements are extracted, but comparison is achieved with respect to a theoretical maximum as well. Both values are measured in Kbps. The deployment scenarios consider an overlapping area created by three BSs as shown in Fig. 1. As the above figure indicates all mobile stations commence their communication, at location points without interference (i.e. outside the overlapping area) while at 1/3 of the data creation duration all nodes rapidly move towards the overlapping area. Consequently 33% of the packet creation duration occurs at no interfering conditions while the rest when residing in the overlapping area. This value is important for the evaluation of the following measurement graphs. It is noted that the specific simulation scenario is selected since it depicts a commonly encountered situation where stations not affected by interference, at same point become affected. In such case the question raised is, on one hand, how the network reacts and, on the other, how effectively an interference minimization technique operates when such a precaution measure is utilized. Thus this particular scenario is selected since it is met in all interference affected scenarios which substantially increases its significance. In order to focus on most valuable conclusions selective measurements are presented. Therefore, the presentation commences with low creation rate per flow with respective measurements presented in Figs. 2 and 3. Measurements made considering such low traffic scenarios while assuming no interference conditions, indicate that all traffic created is successfully transmitted resulting into a throughput ratio equal to 100%. This is the starting point of the presentation since it represents the most relaxed traffic scenario as indicated by delay measurements which remain constant with respect to participating MSs as well as at the lower levels recorded due to zero additional queue delay overhead (Fig. 3). However, the most interesting results are provided by throughput/traffic percentage measurements. This is because, as shown in Fig. 2, up to mid number of MSs (number of

C.P. Antonopoulos / Computer Networks 55 (2011) 356–369

359

Movement at 1/3 of the data creation duration towards overlapping area

BS 2

BS1

BS3

Fig. 1. Three interfering BSs deployment scenario.

Mean Throughput Percentage per BS & 12Kbps Traffic Rate (Throughput / traffic rate) % ratio

100 90 1st BS

80

2nd BS

70

3rd BS

60 50 40 30 20 10 0

2

3

4

5

6

7

8

9 10 11 12 13 14 15 16 17 18 19 20 21 # of Participating MSs

Fig. 2. Arbitration behavior/performance assuming 3 BSs & low data rate.

participating MSs lower than 11) in almost all cases, only two out of the three BSs managed to schedule their communication tasks (respective ratio reaches 100%) at the expense of the third one that was completely overtaken (respective ratio indicated approximately 35%). However, for higher number of MSs resulting to increased aggregate workload, apart from one exception (MSs equal to 15), only one BS was able to schedule successfully its communication tasks while the other two remained to 33% percent-

age. Furthermore, it was observed that in certain cases (13, 14 and 21 participating MSs) BSs’ communications canceled each other out. The reason for this behavior, as extracted from the simulation trace file and code execution analysis, can be attributed to loss of bandwidth request packets losses due to, firstly collisions occurring during the congestion period and secondly because in certain cases MSs got synchronized with the frame of the wrong BSs therefore the frame sequence was disrupted.

360

C.P. Antonopoulos / Computer Networks 55 (2011) 356–369

Mean packet delay per BS & 12 Kbps Traffic Rate 0.0120

Mean Packet Delay (sec)

0.0100 0.0080 0.0060 Poly. (1st BS) Poly. (2nd BS)

0.0040

Poly. (3rd BS)

0.0020 0.0000 0

5

10

15

20

25

# of Participating MSs Fig. 3. Delay behavior/performance assuming 3 BSs and low data rate.

Analogous conclusions are extracted from mid creation rate cases thus not providing additional conclusions, in contrast to excessive traffic scenarios, with respective measurements presented in Figs. 4 and 5. In this case the network is led to saturation point as indicated by the high increase rate of mean delay with values exceeding half a second even from mid number of participating MSs. Additionally thus delay increase is accompanied by significant throughput/traffic ratio decrease exhibited by all BSs. It is noted that this decrease is observed even for BSs that recorded the highest throughput/rate ratio among three BSs at a specific simulation scenario. From these measurements it can be extracted that excessive traffic can significantly alter the network behavior concerning communication over overlapping area. As indicated in specific cases concerning the number of participating MSs (10, 14, 20) the BSs that exhibited higher throughput/traffic ratios (i.e. managed to transmit more data packets) are dif-

ferent compared to the ones observed in relaxed traffic, respective figures. Additionally, especially when the number of participating MSs is higher than 10, the limited channel capacity emerged as the primary performance degradation factor, thus mitigating the interference’s significance. More specifically, focusing on BSs recording the highest throughput measurements, for participating MSs cases 10–12 an almost 10% loss is observed which increases to 25% loss for 14 and 15 participating MSs cases and continues to increase so as to reach 50% and 70% loss for 20 and 21 participating MSs scenario respectively. Additionally a single case is observed (i.e. 13 participating MSs) where the uplink data flows’ interference suppressed each other out. Summarizing, by simulation analysis based on the NS-2 WiMAX model it is shown that concurrently transmitting MSs that reside in overlapping coverage area can lead to significant performance degradation. This degradation is

Mean Throughput Percentage per BS & 1200Kbps Traffic Ratel (Throughput / traffic rate) % ratio

100 90

1st BS

80

2nd BS

70

3rd BS

60 50 40 30 20 10 0 2

3

4

5

6

7

8

9 10 11 12 13 14 15 16 17 18 19 20 21 # of Participating MSs

Fig. 4. Arbitration behavior/performance assuming 3 BSs & high data rate.

361

C.P. Antonopoulos / Computer Networks 55 (2011) 356–369

Mean packet delay per BS & 1200 Kbps Traffic Rate

1.00

Mean Packet Delay (sec)

0.90 0.80 Poly. (2nd BS)

0.70

Poly. (1st BS)

0.60

Poly. (3rd BS)

0.50 0.40 0.30 0.20 0.10 0.00 0

5

10

15

20

25

# of Participating MSs Fig. 5. Delay behavior/performance assuming 3 BSs & high data rate.

observed in both throughput and delay measurements. Furthermore throughput measurements’ patterns are presented and analyzed with respect to each particular BS (since measurements concern the receiving BS), also revealing network performance degradation. This analysis in essence defines the goal of the coordination approach to be proposed, implemented and evaluated. The objectives of this coordinator is mitigating the issue of unfair channel distribution among contending stations thus significantly increasing channel QoS and aggregate throughput of the network. 3. The proposed IP layer BS/BS coordination mechanism In order to tackle the problem of suppressing specific stations access to the transmission medium due to cochannel interference, a controlling unit is proposed. Such an entity should, on one hand be aware of communications causing interference to specific BSs and on the other be able to coordinate BS point to multipoint scheduling towards MSs residing in the overlapping area. Thus, a graphical representation of the system is shown in Fig. 6 where the controlling element is aware of areas either causing or being interference free. In this way the controlling unit is able to instruct which BS is allowed to schedule transmissions towards/from assigned MSs, residing in the overleaping area, during a specific scheduling period. At the same time the rest of competing BSs are inhibited from scheduling data transferring grants from their respective MSs residing in the interference area. The scheduling period is in the order of a few frames thus justifying the characterization of non-real time coordination. Without a doubt, the scheduling period is a critical parameter since it controls the trade-off between scheduling overhead and scheduler responsiveness. Aiming to present a practical and efficient approach, a critical issue is assuring the controlling unit implementation feasibility through the use of elements and entities already defined in the WiMAX working groups able to

assume the role and support the functionality of such a unit. For this reason we refer to [13,14]. WiMAX overview, [13] presents in a consolidated manner the main goals of the WiMAX network working group and offers valuable information for critical issues such as features, implementation scenarios usages modes QoS, mobility management and security. All the above are parts of the network reference model as presented in Fig. 7. From this figure the critical role of the access service network gateway (ASN-GW) element is indicated as well as R6 and R8 reference points describing the communication capabilities among BSs (R8) and between the ASN-GW and BSs (R6). In fact the ASN represents a conceptual entity operating as a logical boundary and representing a convenient way to describe aggregation of functional entities and corresponding message flows associated with the access services [14]. As defined a BS is associated with exactly one default ASN gateway (ASN-GW). A critical aspect advocating the use of ASN-GW as host of the proposed scheduler is indicated by the ASN-GW decomposition which results into two groups of functions, the Decisions Point (DP) and the Enforcement Point (EP). Therefore, such an unit has the role of making decisions as well as enforcing them. Of critical importance are the interfaces also known as reference points and specifically reference point R6 and R8. R6 consists of a set of controls and protocols for communication between the BS and the ASN GW. These include intraASN data path or inter-ASN tunnels between the BS and ASN GW. The control plane, on the other hand, includes protocols for IP tunnel in accordance with the MS mobility events. R6 may also serve as a conduit for exchange of MAC states information between neighboring BSs. R8 reference point defined a set of control plane messages and in some situations, bearer plane data messages between the base stations to ensure fast and seamless handover. Bearer plane consists of protocols that allow data transfer between BSs involved in handover of certain MS. Control plane consists of the inter-BS communication protocol defined in IEEE 802.16 and additional set of protocols that

362

C.P. Antonopoulos / Computer Networks 55 (2011) 356–369

Fig. 6. Indicative example of controlling unit participation.

Fig. 7. WiMAX network reference architecture [13].

allow controlling the data transfer between the Base Stations while an MS performs handover between them. Therefore two options can be suggested concerning the controlling unit implementation.  Use of the ASN-GW as the network device hosting the scheduler making the processing and enforcing the decisions. Communication with BS is achieved through R6 interface messages.  One of the competing BSs can be assigned with the role of hosting the scheduler module. Inter-BS communica-

tion is defined through R8 interface and BS-ASN GW communication can assist through R6 interface as described previously. In both implementation scenarios the mechanism can be divided into two sections. First section includes MSs notifying respective BSs about interference caused by other BSs. Several mechanisms and message structure that could facilitate such notification are proposed in respective WiMAX task groups [15–17]. This information is passed onto the controlling unit using either R6 or R8 messages

363

C.P. Antonopoulos / Computer Networks 55 (2011) 356–369

which is then able to populate appropriate lists indicating which MSs cause interference to which BSs while being associated to which BSs. In this way the controlling unit has all necessary information to specify which BS is allowed transmit during a specific scheduling period and which should not schedule respective uplink flows concurrently. Additionally, if BSs have beam-forming capabilities, enhanced algorithms can be devised taking the antenna pattern parameter into consideration. An overall graphical presentation of this scheduler’s process sequence is shown in Fig. 8. Second phase functionality includes the algorithm through which acquired information is processed. Assuming beam-forming capable BSs stations, the flowchart shown in Fig. 9 indicates the main functionalities. As shown the first step is prioritization of the BSs under control based on their communication demands according to specific approaches. Indicatively such approaches may follow a simple round robin, random selection, proportional fair etc. algorithms. Then the exclusion lists are formed according to the following approach. Starting from the lowest priority BS, for each associated MS it is evaluated whether allowing data transfer from it causes interference to any BS assigned with higher priority. Through the depicted iterative process the communication tasks of a higher priority BS effectively poses restrictions to a lower priority BS. Therefore the highest priority BS facilitates all communication requests while the lowest priority BS accommodates only those that do not compromise any other possible communication. As a last step the controlling unit multicasts the formed exclusion list to controlling BSs for the subsequent scheduling period.

Controlling Entity

BS/RS

Prioritize BS/RS based on specific approaches: 1) Random 2) Round Robin 3) Proportional Fair 4) ???

Does communicating with specific SS causes interference to higher priority BS/RS?

YES

For lowest to highest priority BS/RS

For each SS associated to specific BS/RS

Add specific SS to exclusion list

BS/RS

Control phase start List Containing Interference Causing MSs

Process information Return exclusion lists for each BS/RS

Output exclusion lists List Containing Interference Causing MSs

Process information Return exclusion lists for each BS/RS SS exclusion list

SS exclusion list

Fig. 9. Processing of the acquired interference information.

DATA Phase Based on exclusion lists

Control phase start

Fig. 8. Overall scheduler functionality.

It is noted that the triggering period of this process is of outmost importance along with the chosen prioritization approach. Configuring a too rapid scheduler execution can lead to instability (especially in traffic demanding conditions) since each BS may not have sufficient time to schedule and carry out substantial number of communication events. Additionally, when the network complexity increases this iterative process may require some additional delay. In such cases and assuming a rapid scheduler execution time represented by scheduler execution may become comparable to data transfer period leading to suboptimal

364

C.P. Antonopoulos / Computer Networks 55 (2011) 356–369

network behavior. On the other hand a too slow scheduler may lead to significant unfairness and possibly performance degradation since a specific BS may receive low priority ranking for long periods and packets be lost due to queue overflow or excessive queue delay overhead. 4. Simulation based evaluation of proposed BS coordination mechanism The simulation experiments conducted correspond to scenarios shown in Section 2. A so as to provide immediate correlation and comparison. An additional parameter considered is the scheduling period which varied from 3 up to 10 times the frame duration. The performance of the scheduler varied significantly with respect to this parameter and measurements presented in this section assumed

the optimum network performance recorded so as to provide proof of concept of the proposed approach. Therefore, assuming low creation rate (i.e. equal to 12 Kbps per flow) the beneficial effect of the coordination approach is quite apparent as shown in Figs. 10 and 11. From the above figures it is clearly shown that no channel capturing phenomenon is observed since all three BSs, and in all cases of participating MSs, were able to achieve mean throughput equal to the offered data traffic (respective ratio > 90%). Such behavior indicates a near optimum bandwidth distribution. Furthermore, mean packet delay exhibits a slow increase tendency. Specifically, compared to the evaluation without coordination an average 20 ms increased can be observed which is attributed to the unavoidable scheduling overhead. This increase can be considered acceptable with respect to capacity and

Mean Throughput Percentage per BS & 12 Kbps Traffic Rate (Throughput / Traffic Rate)% ratio

100 90 1st BS

80

2nd BS

70

3rd BS

60 50 40 30 20 10 0 2

3

4

5

6

7

8

9 10 11 12 13 14 15 16 17 18 19 20 21 # of Participating MSs

Fig. 10. Coordination effect on arbitration behavior/performance assuming 3 BSs & low data rate.

Mean packet delay per BS & 12 Kbps Traffic Rate 0.0500 0.0450 Mean Packet Delay (sec)

0.0400 0.0350 0.0300 0.0250 Poly. (1st BS)

0.0200

Poly. (2nd BS)

0.0150

Poly. (3rd BS)

0.0100 0.0050 0.0000 0

5

10

15

20

# of Participating MSs Fig. 11. Coordination effect on delay behavior/performance assuming 3 BSs & low data rate.

25

365

C.P. Antonopoulos / Computer Networks 55 (2011) 356–369

Mean Throughput Transmission Percentage for 3 BSs @ 12 Kbps Traffic Rate

improv default

(Throughput / Traffic Rate) % ratio

100 90 80 70 60 50 40 30 20 10 0 3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

# of Participating MSs Fig. 12. IP coordination effect on aggregate throughput @ low creation rate.

index is bounded between zero and unity and given by Eq. (1). In the context of this equation it is assumed that there are n contenting stations for a specific resource and each one is allocated xi of the resource (xi P 0). If respective allocation is equal to all contenting stations fairness index Findex is equal to 1 indicating an optimally fair system, otherwise the more significant the allocation differences are the lower the index indicating a respectively unfair resource allocation [19,20].

medium allocation enhancement while could be mitigated with scheduler fine tuning. In order to provide a more detailed presentation of the anticipated performance benefits, measurements concerning aggregate throughput increase as well as the channel distribution fairness enhancement are presented. From these figures it is indicated that performance enhancement is rather substantial in all cases. As Fig. 12 indicates, due to the coordinator effect the aggregate throughput achieves a minimum 20% increase which reaches up to 40% and 60% in several simulation scenarios. Focusing on fairness, on the other hand, two types of graphs are presented. The first presents the average differences of throughput/traffic ratios regarding flows received by the three competing BSs. Such representation offers fairness perception through absolute throughput measurements. On the other hand, the second type presents fairness through the well established and widely used Jain’s index metric [18]. As indicated in respective literature this

Pn F index ¼

2 i¼1 xi P n ni¼1 x2i

ð1Þ

:

The fact that in most cases at least one BS suffers sever throughput degradation without the use of coordination has a clearly negative effect on bandwidth fairness distribution as indicated in Fig. 13. In this figure it is depicted by the ‘‘default” columns a significant 40% throughput/ traffic ratio difference amongst contenting BSs in almost

Average inter-BS % Throughput/Traffic difference for 3 BSs @ 12Kbps Traffic Rate (Throughput / Traffic Rate) % ratio difference

70 improv

60

default

50 40 30 20

10 0 2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

# of Participating MSs Fig. 13. IP coordination effect on throughput allocation @ low creation rate.

19

20

21

366

C.P. Antonopoulos / Computer Networks 55 (2011) 356–369

Throughput/Traffic Rate Fairness for 3 BSs @ 12Kbps Traffic Rate

improv default

1 0.9

Jain's Fairness index

0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

# of Participating MSs Fig. 14. IP coordination effect on Jain’s fairness index @ low creation rate.

Mean Throughput Transmission Percentage for 3 BSs @ 1200 Kbps Traffic Rate (Throughput / Traffic Rate) % ratio

100 improv

90

default

80 70 60 50 40 30 20 10 0 3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

# of Participating MSs

Fig. 15. IP coordination effect on aggregated throughput @ high creation rate.

all cases. However, using the coordinating approach respective measurement did not exceed 5%. Therefore, it is shown that the presented coordination approach did in fact manage to fairly distribute the available bandwidth since the difference of achieved throughput with respect to imposed traffic is almost negligible. Measurements of Jain’s fairness index also verify the positive effect of the proposed coordination approach as presented in Fig. 14. As indicated in all cases a more than 10% increase is observed, comparing coordinated as opposed to default network scenarios, which reaches up to 22% in several cases (6, 9, 12, 16, 17, 18, 19, 20 # of participating MSs). However, it is noted that in order to extract objective conclusions, such measurements should be analyzed in conjunction with aggregate throughput respective ones. For example, in 13 and 14 participating MSs cases, Jain’s index may indicate that both the default and the coordinated network offer optimum fairness, but for the same cases Fig. 12 indicates a significant aggregate throughput favoring the coordinator functionality.

Concerning high packet creation rate (i.e. 1200 Kbps per flow) only graphs concerning comparative evaluation of network aggregate throughput and bandwidth distribution fairness are presented (Figs. 15 and 16). This is because respective throughput/traffic ratio and delay related figures offer conclusions of low significance. On one hand, mean throughput/traffic graphs presents a behavior similar to the default network’s patterns. This is attributed to the limited capacity factor drastically affecting network behavior invalidating or at least minimizing the effect of the coordinating algorithm. On the other hand, delay graphs are again analogous to those considering the default network indicating the performance degradation can be solely correlated to exceeded limited network capacity. Thus is not possible to extract safe conclusions concerning the effect of transmitting over the common transmission area. However, Figs. 15–17 do provide additional useful insights concerning the coordination mechanism behavior. Thus from Fig. 15 it is indicated that, focusing on aggregate

367

C.P. Antonopoulos / Computer Networks 55 (2011) 356–369

Average inter-BS % Throughput/Traffic difference for 3 BSs @ 1200Kbps Traffic Rate

(Throughput / Traffic Rate) % ratio difference

70

improv

60

default

50 40 30 20 10 0 # of Participating MSs

Fig. 16. IP coordination effect bandwidth distribution fairness @ high creation rate.

Throughput/Traffic Rate Fairness for 3 BSs @ 1200Kbps Traffic Rate

improv default

1

Jain's Fairness index

0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

# of Participating MSs Fig. 17. IP coordination effect on Jain’s fairness index @ low creation rate.

achieved throughput/traffic ratio, in excessive traffic situations differences between the network with and without the coordination approach are marginal favoring coordination approach for up to 10 participating MSs. However, the most important conclusion can be extracted from Figs. 16 and 17 which once again advocate the use of the coordination mechanism when aiming towards optimum bandwidth allocation. As observed, compared to Fig. 13 measurements, the margin between the default and the coordinated network is considerably reduced. However the coordination approach based differences are still substantially lower compared to the ones extracted by the ‘‘default” network. Hence, even in excessive traffic scenarios the coordinator managed to reduce achieved throughput differences more than 50% compared to the default network. However, it is noted that in specific scenarios (participating MSs equal to 10, 11, 12, 15) where the benefits are rather marginal.

Additional conclusions can be extracted from the Jain’s index measurements. As indicated the coordinated network, in several cases, fails to offer optimum bandwidth distribution although providing enhanced performance compared to the default network in all but one case. More specifically through the use of the coordination approach Jain’s index is always higher than 0.9, thus being 10% to 20% higher compared to the default network’s performance in most cases. However, it is noted that in two cases it is recorded the default network being fairer than the coordinated, indicative of the coordinator’s difficulty to apply its control under congested scenarios. Summarizing, it is shown that the presented coordination approach offers performance benefits on all cases. Focusing on scenarios where traffic is not excessive both aggregate throughput with respect to imposed traffic exhibited significant increase while the transmission channel was optimally distrusted among all data flows. On the

368

C.P. Antonopoulos / Computer Networks 55 (2011) 356–369

other hand, marginal unavoidable delay increase is observed due to the coordinator algorithm effect. Furthermore, focusing on excessive traffic scenarios the coordinator approach offers considerable aggregate throughput increase and most importantly significant enhancement as far as the channel distribution is concerned. Thus, in all cases the presented approach provided behavior enhancement both quantitatively as well as qualitatively.

5. Conclusions Interference due to adjacent co-channel BS transmission constitutes a critical issue preventing optimum resource utilization in WiMAX networks while causing severe network performance degradation. Based on this observation, the main objective of this paper is twofold. Firstly, evaluate the significance of this issue in the context of a well known system level network simulator, while secondly propose an efficient, feasible and of low complexity coordination mechanism able to alleviate negative effect of this predicament. Respective proposition is evaluated in terms of bandwidth optimum distribution and aggregate throughput based on a well known and widely used simulation platform. In this context, evaluation concerning the network behavior without the assistance of the proposed coordination mechanism, varying a wide range of network parameters, proved that transmission over overlapping coverage areas, assuming co-channel communication, leads to severely unfair channel distribution, thus compromising network capacity performance and robustness. Furthermore it has provided valuable information regarding network behavior in respective simulation scenarios. Driven by the aforementioned evaluation’s conclusions, an efficient coordination mechanism is specified enabling, on one hand, interference related information acquisition and, on the other, adequate processing assuring prioritized time scheduling. Additionally, as an added value, concerning the specifications of the proposed approach emphasis has being given on issues such as implementation feasibility and low complexity, which are of critical importance, especially to the industrial community. Respectively the presentation of the approach is based on the use of already existing entities as well as sets of messages through respective reference points of well known network reference models. Simulation based evaluation, using the NS2 network simulator platform, of the presented mechanism revealed significant performance enhancement in all considered scenarios. In scenarios assuming relaxed data traffic demands, system performance and behavior benefits are most apparent since aggregate achieved throughput over common communication areas is significantly compensated and thus BSs are able to transmit with minimum losses. At same time optimally fair channel distribution is observed since average throughput differences amongst BSs are always below 5% and respective Jain’s index measurements always higher than 0.99. When aggregate workload reaches saturation point, limited channel capacity factor emerges as the main reason for performance degradation. Hence, with respect to the

achieved throughput, the proposed approach offers only marginal difference from network scenarios without coordination. However, even in such stressed conditions the coordination algorithm is shown to provide considerably fairer channel distribution as indicated by respective measurements. Concluding, this paper presents an approach able to increase network performance in wide range application scenarios, while imposing acceptable overhead. This approach is based on IP layer communication following a perspective indicating that, respective mechanisms should aim toward system level performance enhancements, through realistic and feasible solutions from an implementation point of view. Such features increase considerably the added value and effective usefulness of respective proposals. Additionally, it comprises a valuable basis for future extensions covering specific or enhanced application and network demands. Acknowledgements This work has been partially performed within the framework of the ROCKET ICT-215282 FP7 project that is co-funded by the European Commission References [1] IEEE 802.16j TG, Part 16: Air interface for fixed and mobile broadband wireless access systems multihop relay specification, Baseline Document for Draft Standard for Local and Metropolitan Area Networks, 6/6/2007. [2] C. Hoymann, Analysis and performance evaluation of the OFDMbased metropolitan area network IEEE 820.16, Computer Networks 49 (2005) 341–363. [3] F. Rashid-Farrokhi, K.J. Ray Liu, L. Tassiulas, Transmit beam-forming and power control for cellular wireless systems, IEEE J Sel Area Comm 16 (8) (1998) 1437–1450. [4] Ch. Hoymann, H. Meng, J. Ellenbeck, Influence of SDMA-specific MAC scheduling on the performance of IEEE 802.16 networks, in: 12th European Wireless Conference, April 2–5, 2006, Athens, Greece. [5] M. Ergen, S. Coleri, P. Varaiya, QoS aware adaptive resource allocation techniques for fair scheduling in OFDMA based broadband wireless access systems, IEEE Trans Broadcast 49 (4) (2003) 362–370. [6] Y. Jun Zhang, K. Ben Letaief, Optimizing power and resource management for multiuser MIMO/OFDM Systems, IEEE GLOBECOM 03, 1–5 Dec. 2003, San Francisco, USA. [7] J. Zou, D. Zhao, Real-time CBR traffic scheduling in IEEE 802.16-based wireless mesh networks, Springer Wireless Networks 2007, ISSN 1022-0038 (Print) 1572-8196 (Online). [8] M.H. Ahmed, H. Yanikomeroglu, S. Mahmoud, Interference management using base-station coordination in broadband wireless access networks, Wireless Commun Mobile Comput 6 (2006) 95– 103. [9] T. Fong, P. Henry, K. Leung, X. Qiu, N. Shankaranarayanan, Radio resource allocation in fixed broadband wireless networks, IEEE Trans Commun 46 (6) (1998) 806–817. [10] K. Leung, A. Srivastava, Dynamic allocation of downlink and uplink resource for broadband service in fixed wireless networks, IEEE J Sel Area Commun 17 (5) (1999) 990–1006. [11] L. Mailaender, H. Huang, H. Viswanathan, Simple inter-cell coordination schemes for a high speed CDMA packet downlink, in: Proc. IEEE VTC’00 Spring, Tokyo, Japan, May 2002. [12] R. Rouil, The network simulator NS-2 NIST add-on IEEE 802.16 model (MAC+PHY) Draft 1.0, National Institute of Standards and Technology (NIST), Sept. 2006. [13] P. Yegani, WiMAX overview. IETF-64, Nov. 7–11, 2005 Vancouver, Canada, Copyright 2003 Cisco Systems. [14] WiMAX forum network architecture (Stage 2: Architecture tenets, reference model and reference points) Release 1.0.0, March 28, 2007 WiMAX Forum Proprietary, Copyright 2005–2007 WiMAX Forum.

C.P. Antonopoulos / Computer Networks 55 (2011) 356–369 [15] Chun-Ting Chou, I-Kang Fu, Paul Cheng, Messaging design to support FFR operation, IEEE 802.16 Broadband Wireless Access Working Group, 27/4/2009. [16] Lei Wang, Proposed MAC messages related to the DL measurement/ report in multicarrier systems (16.2.8.2.8), IEEE 802.16 Broadband Wireless Access Working Group, 31/12/2009. [17] Wei-Peng Chen, Chenxi Zhu, Interference detection and measurement in OFDMA relay networks, IEEE 802.16 Broadband Wireless Access Working Group, 14/3/2007. [18] R. Jain, D. Chiu, W. Hawe, A quantitative measure of fairness and discrimination for resource allocation in shared computer systems, Digital Equipment Corporation, Maynard, Mass, USA, 1984. [19] G. Berger-Sabbate, A. Duda, O. Gaudoin, M. Heusse, F. Rousseau, Fairness and its impact on delay in 802.11 networks, in: Proc. IEEE Global Telecommunications Conf (GLOBECOM ’04), vol. 5, Dallas, Tex, USA, 2004, pp. 2967–2973. [20] F.A. Bokhari, H. Yanikomeroglu, W.K. Wong, M. Rahman, Cross-layer resource scheduling for video traffic in the downlink of OFDMAbased wireless 4G networks, EURASIP J Wireless Commun Network 2009 (2009). Article No. 3, ISSN 1687-1472. [21] http://nsnam.isi.edu/nsnam/index.php/Main_Page.

369

Christos P. Antonopoulos ([email protected]), received his Diploma in Electrical Engineering and Computer Technology and his Ph.D. degree in Electrical Engineering from the Department of Electrical Engineering and Computer Technology, University of Patras, Greece in 2002 and 2008 respectively. From 2002 he is a Research Assistant in the Applied Electronics Laboratory of the same department participating in European and National research projects. From August 2006 he is also a researcher for INTRACOM–TELECOM a lead telecommunication company in Greece, in the context of European research projects. His main research interests include Wireless Communication, Wireless LANs, Ad-Hoc wireless networks, Sensor networks, Cross-layer design, MAC layer protocols, Wireless metropolitan area networks.