利用可動態轉向之有向性天線改善IEEE 802.16(d)網狀網路之效能
Using Steerable Directional Antennas to Improve
the Performances of IEEE 802.16(d) Mesh Networks
研 究
生：許登偉
Student
：Teng-Wei Hsu
指導教授：王協源
Advisor
：Shie-Yuan Wang
國 立 交 通 大 學
網 路
工 程 研 究 所
碩
士 論 文
A Thesis
Submitted to Institute of Network Engineering College of Computer Science
National Chiao Tung University in partial Fulfillment of the Requirements
for the Degree of Master
in
Computer Science
June 2007
Hsinchu, Taiwan, Republic of China
利用可動態轉向之有向性天線改善IEEE 802.16(d)網狀網路之效能
Using Steerable Directional Antennas to Improve
the Performances of IEEE 802.16(d) Mesh Networks
研究生：許登偉 指導教授：王協源 國 立 交 通 大 學 網 路 工 程 研 究 所 摘要 IEEE 802.16(d)網狀網路很有希望成為下一代的無線骨幹網路，在該網路中 的節點皆須配備一隻全向性天線（或是由多隻有向性天線所構成的天線陣列） 來送收無線訊號。在此網路下，有效發揮有向性天線空間重複利用的特性將會 是一個很大的挑戰。 在本篇論文中，相對於原本使用多隻有向性天線所構成的天線陣列去模仿全 向性天線的功能，我們提出了一個全新的設計，使得每個網路節點只須配備一 隻可動態轉向的有向性天線，我們的設計可以更佳的發揮有向性天線的優勢並 且大量減少網路部署的成本。
由我們在媒體存取控制（Media Access Control）層的模擬結果，可以觀察 到我們的設計有效的提升可使用之控制頻寬的利用率。另一方面，應用程式所 能得到的傳輸流量顯示，我們的設計可以提升整體網路TCP傳輸流量7.506倍以 及UDP傳輸流量2.436倍。 關 關 關鍵鍵鍵字字字：：：有有有向向向性性性天天天線線線、、、網網網狀狀狀網網網路路路、、、無無無線線線都都都會會會區區區域域域網網網路路路、、、IEEE 802.16、、、WiMAX。。。
Using Steerable Directional Antennas to Improve
the Performances of IEEE 802.16(d) Mesh Networks
Student: Teng-Wei Hsu Advisor: Prof. Shie-Yuan Wang
Institute of Network Engineering National Chiao Tung University
Abstract
The IEEE 802.16(d) mesh network is a promising next-generation wireless back-bone network. This network requires that all network nodes should be equipped with an omni-directional antenna (or an directional antenna array emulating it). Exploiting the spatial reuse property of directional antennas in such networks is a great challenge.
In this thesis, instead of collocating directional antennas to emulate omni-directional antennas, we propose a novel design using only one steerable omni-directional antenna for each node. Our design can better exploit the advantages of directional antennas and greatly reduce the deployment cost of the network.
Our MAC-layer simulation results show that our design significantly increases the utilization of the control-plane bandwidth. The application throughput results show that our design can increase the aggregate TCP throughputs by a factor of 7.506 and the aggregate UDP throughputs by a factor of 2.436, respectively.
Keywords: directional antenna, mesh network, wireless metropolitan area net-works, IEEE 802.16, WiMAX.
致 致 致謝謝謝辭辭辭 我我我能能能夠夠夠擁擁有擁有有充充充實實實的的的研研研究究究生生生生生生涯涯涯，，，要要要感感感謝謝謝我我我的的的指指指導導導教教教授授授王王王協協協源源源老老老師師師，，，提提提供供供給給給 我 我 我一一一個個個能能能充充充分分分訓訓訓練練練且加且且加加強強強研研研究究能究能能力力力與與與實實實務務務經經經驗驗驗的的的環環環境境境，，，並並並在在在兩兩兩年年年之之之間間間與與與我我我分分分 享 享 享其其其多多多年年年的的的研研研究究究與與與實實務實務務經經經驗驗驗，，，幫幫幫助助助我我我在在在關關關鍵鍵鍵的的的時時時刻刻刻能能能做做做出出最出最最佳佳佳的的的決決定決定定。。。另另另外外外 要 要 要感感感謝謝謝兩兩兩位位位博博博班班班學學學長長長——周周周智智智良良良與與與林林林志志志哲哲哲，，，在在在學學學長長長的的的帶帶帶領領領之之之下下下，，，讓讓讓我我我擁擁擁有有有團團團 隊 隊 隊合合合作作作以以以及及及設設設計計計與與與維維護維護護大大大型型型專專案專案案的的的經經經驗驗驗，，，並並並且且且在在在撰撰撰寫寫寫論論論文文時文時時給給給我我我許許多許多多寶寶寶貴貴貴的的的 建 建 建議議議，，，因因因此此此我我我深深深切切切的的的期期期許許許學學學長長長能能能與與與我我我分分分享享享附附附與與與本本本篇篇篇論論論文文文的的的榮榮榮耀耀耀。。。此此此外外外我我我要要要感感感 謝 謝 謝能能能撥撥撥冗冗冗擔擔擔任任任本本本篇篇篇論論論文文文口口口試試試委委委員員員的的的三三三位位位教教教授授授，，，你你你們們們珍珍珍貴貴貴的的的建建建議議議讓讓讓我我我的的的論論論文文文能能能 夠 夠 夠更更更加加加的的的充充充實實實且且且完完完整整整。。。 這這這兩兩兩年年年我我我之之之所所所以以以能能在能在在各各各方方方面面面不不不斷斷斷地地地茁茁茁壯壯壯，，，要要要感感感謝謝謝實實實驗驗驗室室室與與與系系系上上上計計計算算算機機機中中中心心心 的 的 的每每每一一一位位位同同伴同伴伴，，，在在在我我我遇遇遇到到到無無無法法法突突突破破破的的的難難難題題題時時時，，，你你你們們們即即即時時時的的的建建建議議議總總總是是是能能能幫幫幫助助助我我我 跨 跨 跨越越越這這這些些些困困困難難難繼繼繼續續續前進前前進進，，，此此此外外外更更更重重重要要的要的的，，，在在在平平平時時時的的的相相相處處處中你中中你你們們們不不不時時的時的的帶帶帶個個個我我我 各 各 各式式式各各各樣樣樣不不同不同同的的的歡歡歡樂樂樂，，，讓讓讓我我我這這這兩兩兩年年年的的的研研研究究究生生生涯涯涯添添添上上上努努努力力力與與嚴與嚴嚴肅肅肅以以以外外外的的的色色色彩彩彩。。。 最最最後後後要要要感感感謝謝謝我我我的的的父父母父母母以以以及及及一一一直直直伴伴伴隨隨隨我我我的的的人人人，，，你你你們們們總總總是是是能能能在在各在各各方方方面面面無無無條條條件件件地地地 支 支 支持持持我我我關關關心心心我我我鼓鼓鼓勵勵勵我我，我，，讓讓讓我我我無無需無需需擔擔擔心心心生生生活活活上上上的的的問問問題題題，，，並並並且且且在在在我我我心心心情情情低低低落落落時時時不不不 厭 厭 厭其其其煩煩煩地地地聆聆聆聽聽聽我我我的的的抱抱怨抱怨怨，，，在在在我我煩我煩煩惱惱惱時時時不不不間間間斷斷斷的的的給給給我我我鼓鼓鼓勵勵勵，，，讓讓讓我我我能能能一一一直直直努努努力力力至至至 今 今 今天天天。。。
Contents
摘 摘 摘要要要 i Abstract ii 致 致 致謝謝謝辭辭辭 iii Contents ivList of Figures vii
List of Tables ix
1 Introduction 1
2 Related Work 3
3 Background 9
3.1 Overview of IEEE 802.16(d) Mesh Networks . . . 9 3.1.1 Node, Neighbor, Neighborhood, and Extended Neighborhood 10 3.1.2 Network Entry . . . 10 3.1.3 Contention Resolution for Transmitting Control Messages . . 11 3.1.4 Distributed Coordinated Data Transmission Scheduling . . . 12 3.2 Dynamic Holdoff Time Setting . . . 13 3.2.1 The Holdoff Time in the IEEE 802.16 Mesh Network . . . . 13 3.2.2 Dynamic Holdoff Time Approach . . . 14
4 Protocol Design 16 4.1 Difficulty of using Directional-antenna in IEEE 802.16 Mesh Network 16
4.2 Introduction to the Omni-direction-antenna MAC layer . . . 17
4.3 Directional-antenna Related Terminologies . . . 20
4.4 Modified Network Entry Process . . . 21
4.4.1 Issues involved in Using Directional-antenna . . . 21
4.4.2 Node Selection for Synchronization and Network Entry . . . 21
4.4.3 Antenna Orientation for MSH-NENT message Transmission 23 4.5 Modified Contention Resolution for Control Messages . . . 24
4.5.1 Issues of using Directional-antenna . . . 24
4.5.2 Antenna Orientation for MSH-NCFG and MSH-DSCH mes-sages Transmission . . . 25
4.5.3 Modifications of Control Message Scheduler . . . 27
4.5.4 Modified Determination of Eligible Nodes within a Node’s Extended Neighborhood . . . 34
4.6 Modified Data Transmission Time Scheduling . . . 38
4.6.1 Validity of Allocation in the Omni-direction-antenna Network 38 4.6.2 Validity of Allocation in the Directional-antenna Network . . 40
5 Performance Evaluation 43 5.1 Simulation Parameters . . . 43 5.2 MAC-layer Performance . . . 44 5.2.1 Performance Metrics . . . 44 5.2.2 Simulation Environment . . . 45 5.2.3 Simulation Results . . . 46 5.3 Application Throughputs . . . 52 5.3.1 Performance Metrics . . . 52 5.3.2 Simulation Environment . . . 52 5.3.3 Simulation Results . . . 53 6 Future Work 64
7 Conclusion 66
List of Figures
3.1 The holdoff time before contending for next transmission opportunity. 14
4.1 Components of the MAC-layer Module . . . 17
4.2 Mesh Frame Structure. . . 18
4.3 Antenna domain definition. . . 20
4.4 The refined mesh election algorithm. . . 22
4.5 Antennae orientation for MSH-NENT transmission. . . 24
4.6 Receiving control messages from the sending node is more difficult in directional-antenna networks. . . 26
4.7 The offset used for expressing the transmission opportunity exactly. 27 4.8 Directional-antenna version Mesh Election Algorithm. . . 28
4.9 Determine the next transmission opportunity for the antenna do-main with index 2 . . . 29
4.10 Using different holdoff time values for different antenna domains. . . 31
4.11 Dynamic Holdoff Time Exponent Determination Algorithm . . . 33
4.12 A comparison between the time required for establishing a data schedule with and without considering dynamic bandwidth needs . 33 4.13 Issue in determining the eligibility of a one-hop neighbor. . . 35
4.14 Multiple next transmission opportunities of the two-hop neighbors. 37 4.15 Data scheduling in the omni-directional-antenna network. . . 40
4.16 Data scheduler in the directional-antenna network. . . 42
5.1 The simulation network topology. . . 45
5.3 ATHPT versus α value. . . 50 5.4 ANEDS versus α value. . . 51 5.5 TCP throughput using antenna with beam width π2 with DEDA. . . 56 5.6 TCP throughput using antenna with beam width π3 with DEDA. . . 57 5.7 TCP throughput using antenna with beam width π2 with
DEDA-ITHP. . . 58 5.8 TCP throughput using antenna with beam width π3 with
DEDA-ITHP. . . 59 5.9 UDP throughput using antenna with beam width π2 with DEDA. . . 60 5.10 UDP throughput using antenna with beam width π3 with DEDA. . . 61 5.11 UDP throughput using antenna with beam width π2 with
DEDA-ITHP. . . 62 5.12 UDP throughput using antenna with beam width π3 with
List of Tables
5.1 MAC-layer results using antenna with beam width π2 . . . 47
5.2 MAC-layer results using antenna with beam width π3 . . . 48
5.3 Application throughputs using antenna with beam width π2 . . . 54
Chapter 1
Introduction
The IEEE 802.16(d) standard (WiMAX) [5] is a promising candidate for next-generation wireless broadband technology. In this standard, two operational modes are defined. One is the point-to-multipoint (PMP) mode, which provides one-hop communications between a base station and several subscriber stations. The other is the mesh mode, which supports multi-hop and peer-to-peer communications. Using the mesh mode, subscriber stations can directly communicate with each other without the aid of the base station.
The directional antenna technology features the well-known spatial reuse prop-erty for wireless signal. It can increase network capacity and data transmission concurrency. The IEEE 802.16(d) standard has defined the adaptive antenna sys-tem (AAS) for error resilience and transmission concurrency in the PMP mode. AAS exploits a set of directional sector antennas to form an antenna array. In con-trast, the mesh mode is defined based on the omni-directional-antenna assumption and operates using broadcast control messages. To emulate the broadcasting func-tion of an omni-direcfunc-tional antenna, the standard allows collocating sector anten-nas to emulate an omni-directional antenna for each network node. In such a way, however, the spatial reuse property of directional antennas cannot be exploited in the mesh mode and therefore network capacity cannot be substantially increased. In this thesis, we propose a novel design to better exploit the advantages of directional antennas for the IEEE 802.16 mesh network. The proposed design
is based on the mesh distributed coordinated scheduling (MSH-DSCH) mode and only requires that each network node be only equipped with a steerable directional antenna. This design employs our proposed innovative management processes, which are extended from the original standard, to eliminate the need of emulating control message broadcasting.
Such a novel design has two advantages. First, compared with the sector antenna array design, the deployment cost of our design can be much reduced because using our design each network node is equipped with only one directional antenna. Second, the spatial reuse property can be efficiently exploited. Thus, the capacity and data transmission concurrency of a network can be much increased. To the best of our knowledge, in the literature no papers have studied how to use steerable directional antennas for IEEE 802.16(d) mesh networks. As such, our thesis has two contributions. First, this thesis is the first paper proposing a novel design to make the steerable-directional-antenna system feasible in an IEEE 802.16(d) mesh network. Second, the proposed design can much increase network capacity and data transmission concurrency for the IEEE 802.16(d) mesh network. The rest of this thesis is organized as follows. In Chapter 3, an overview of the IEEE 802.16 mesh network are introduced. The preliminary fundamentals and re-lated work are also presented. In Chapter 4, we elaborate on our proposed design. In Chapter 5, we compare the performances of the IEEE 802.16(d) mesh networks using our design and the original omni-directional-antenna design. In Chapter 6, advanced issues and improvements for our design are addressed. Finally, in Chap-ter 7 we conclude this thesis.
Chapter 2
Related Work
In recent years, using directional antennas to increase capacity and spatial reuse of wireless networks has been extensively studied. However, most of previous studies are based on IEEE 802.11(b) networks. These papers deal with issues of using directional antennas in a CSMA/CA based networks, such as neighboring node discovery, antenna orientation, and routing. To the best of our knowledge, our work is the first paper studying issues of using steerable directional antennas in IEEE 802.16(d) mesh networks.
In [4], the authors propose a new network architecture with multi-channel multi-sector directional antennas (MCMSDA WLAN). Based on the proposed ar-chitecture, they also propose a Lagrangian Relaxation based algorithm for balanc-ing loads in the vicinity of several neighborbalanc-ing access points. This work deals with the load-balancing problem for links in an IEEE 802.11(b) WLAN; thus, the scope of this paper is far from that of ours.
In contrast to WLANs, using directional antennas in ad hoc networks is much more challenging, requiring additional mechanisms to discover potential neighbor-ing nodes and coordinate antenna orientation. In addition to these two essential problem, “deafness” is another problem that greatly decrease the performance gain when using directional antennas. [19],[2], [9], [6], [7], [13], and [14] propose modifications to the IEEE 802.11(b) MAC protocol to address the above issues. These variants of the IEEE 802.11(b) MAC standard are CSMA/CA based
pro-tocols and, therefore, much differ from our work, which aims to using directional antennas in IEEE 802.16(d) mesh networks.
In [19], the authors modify the original IEEE 802.11(b) virtual carrier sens-ing mechanism that utilizes the RTS and CTS control messages. They present a directional virtual carrier sensing mechanism (DVCS), in which MAC frames (RTS, CTS, DATA, and ACK frames) are directionally transmitted. The DVCS mechanism records the angle-of-arrival information of incoming packets sent by neighboring nodes to optimally adjust the orientation of antennas when transmit-ting packets back to those nodes.
The DVCS mechanism also modifies the network allocation vector control mes-sage (NAV mesmes-sage), defined by the IEEE 802.11(b) standard, to be a directional-antenna-version variant (DNAV message). The NAV message is used to inform neighboring nodes of a duration of forthcoming packet transmission, during which nodes other than the transmitting node should suspend their packet transmission to avoid packet collisions. The NAV message is designed to be omni-directionally transmitted while the DNAV message can be directionally transmitted. As a re-sult, the DNAV message can reduce the number of neighboring nodes that should suspend their data transmission, as compared with the NAV message, therefore increasing the spatial reuse level of a network.
[7] proposes a modified IEEE 802.11(b) MAC protocol, named directional MAC (DMAC). This version of DMAC comprises two schemes. Using the first scheme, a transmitting node transmits RTS messages directionally (DRTS) to the intended receiving node. On receiving the DRTS message, the receiving node should omni-directionally send a CTS message to notify its neighboring nodes of the forth-coming packet reception, if the channel is clear for packet reception. Using the second scheme, a transmitting node can send RTS messages either directionally or omni-directionally depending on two rules. (1) if any antennas of the transmitting node and its neighboring nodes are suspended due to the CSMA/CA mechanism (for example, one neighboring node is going to transmit or receive packets.), this transmitting node should directionally send its RTS message to the intended
re-ceiving node. (2) In other cases, the transmitting node should omni-diretionally transmit its RTS message. The use of DRTS increases the spatial reuse degree of a wireless network and thus increases the capacity of the network.
In [2] and [14], Choudhury et al. propose another version of directional MAC protocol for exploiting the advantages of using directional antennas in IEEE 802.11(b) networks. This version of DMAC directionally transmits all types of MAC frames to increase spatial reuse of an IEEE 802.11(b) network and allows a multi-hop RTS operation to establish multi-hop links between distant nodes (The DMAC with multi-hop RTS operation is referred to as MMAC in [2].). Using the multi-hop RTS operation, a transmitting node can initiate a multi-hop RTS frame destined to a node several hops away from it. This multi-hop RTS frame can be relayed by intermediate nodes and set up a direct packet transmission from the transmitting node to the receiving node. As a result, MMAC can reduce the number of packet transmissions for multi-hop links, as compared to DMAC.
On the other hand, to solve the deafness problem, for each transmission pair (a transmitting and a receiving nodes), [7] categorizes other transmission pair into two types: related and unrelated traffic to it. Related traffic is defined as the set of transmission pairs that may interfere with each other. In the MAC protocol proposed in [7], a transmitting node of a transmission pair should adjust its antenna beamwidth to cover nodes that are proceeding data packet transmission/reception. In such a design, nodes belonging to the same related traffic set can be aware of data packet transmission/reception that may interfere with their own traffic. As such, data packet collisions can be avoided with the original CSMA/CA MAC mechanism.
Choudhury and Vaidya propose a tone-based MAC protocol (ToneMAC) to solve the deafness problem in [13]. Besides the original CSMA/CA protocol, ToneMAC uses an additional out-of-band tone signal to help network nodes dif-ferentiate transmission failures due to collisions from those due to deafness. Using ToneMAC, RTS, CTS, DATA, and ACK frames are transmitted directionally to exploit the spatial reuse advantage of directional antennas. Besides, a receiving
node should omni-directionally transmit an out-of-band tone signal after it sends an ACK frame out and, similarly, a transmitting node should omni-directionally transmit the tone signal after it receives an ACK frame acknowledging the DATA frame it has transmitted. For nodes neighboring to transmitting and receiving nodes, they can tell MAC frame collisions from the node deafness condition. If they receive tone signals from its intended receiving node, they will know that the intended receiving node pointed its antenna to another direction for data trans-mission/reception, thus causing a deafness condition.
In [9], Ramanathan et al. summarize issues of using directional antennas in IEEE 802.11(b) networks for the data link layer, MAC layer, and routing proto-col. They also propose a generic frame for neighboring node discovery and power control in [9].
Regarding routing protocol for directiona-antenna networks, [12] proposes a new CSMA/CA based MAC protocol with the aid of topology and packet transmis-sion information. Before starting data frame transmistransmis-sion/reception directionally, each transmitting/receiving node should omni-directionally transmit RTS/CTS frames to help its neighboring nodes keep track of the information of topology and on-going communication pair. Such a MAC protocol design enables use of antenna-pattern-aware routing protocol for better load-balancing, as compared to previously proposed routing protocols. [12] also proposes such a antenna-pattern-aware protocol to find routing paths that can minimize the interference with other communication pairs, thereby effectively balancing network loads over all network nodes.
Besides proposing modifications and enhancements to the IEEE 802.11(b) MAC protocol, several works have proposed alternative MAC protocols for wireless ad hoc networks. In [15], the authors present a slotted-aloha-based MAC protocol with adaptive array smart antennas. The performance evaluation is carried out in both analytical and simulation approaches. The performances of the proposed slotted-aloha protocol are studied using varied number of antenna elements and network loads in terms of throughputs and packet delay. Performance comparison
between the proposed protocol and the original IEEE 802.11(b) is also given in this work.
In [10], Raman and Chebrolu design and implement the 2P MAC protocol to replace the existent IEEE 802.11 CSMA/CA protocol in the context of wireless mesh networks. The 2P MAC protocol uses a token-based approach to coordinate network nodes’ transmission and reception. With the aid of expiration timers, 2P is capable of synchronizing network nodes for successful packet transmission and reception. The authors build a real-world testbed to compare the performances of the 2P and CSMA/CA protocols in terms of UDP and TCP throughputs. The experiment results show that 2P outperforms the CSMA/CA protocol regarding application throughputs.
In [8], Navda et al. design and implement a beam steering framework us-ing steerable directional antennas (MobiSteer) to improve performances of IEEE 802.11 links between a moving vehicle and roadside access points. The results of field trials show that the link quality (in terms of SNR value) obtained by MobiSteer is better than the network using omni-directional antennas.
Instead of designing or implementing new MAC protocols for directional anten-nas, [11] and [18] evaluate the performance of existing antenna technologies and protocols. In [11], Ramanathan compares the performances of steerable antennas and switched antennas using antenna patterns. Simulations with a realistic radio and propagation model were conducted to study maximum achievable through-puts and delays of networks using these two antennas with varied gains and node density. In [18], Ueda et al. built a real-world testbed for evaluating wireless ad hoc network with smart antennas. They realized and evaluated a spatial division multiple access (SDMA) protocol, which periodically collects the direction and signal level of neighboring nodes, stores them into an angle-signal-table (AST), and determines the used antenna pattern according to the information of AST.
Most of the literature is based on IEEE 802.11(b) CSMA/CA protocol. [1] is one of few papers discussing the IEEE 802.16(d) mesh network using directional antennas. The authors assume that each network node is equipped with directional
antennas and propose an integer programming approach to minimize the deploy-ment cost of an IEEE 802.16-based backhaul network. However, this paper does not consider the coordination overhead of control message transmissions, when using directional antenna in IEEE 802.16(d) mesh networks. Such coordination of control message transmission is essential to solve the deafness problem. The rea-son is that the IEEE 802.16(d) mesh network is a coordinated network; thus, each node can compute the transmission timing of its neighboring nodes. With the aid of this information, a receiving node can point its antenna to the intended trans-mitting node at the correct timing. Nonetheless, the IEEE 802.16(d) mesh-mode MAC protocol coordinates control message transmissions with the assumption of the broadcasting nature of wireless signal. As such, adapting the original IEEE 802.16(d) mesh-mode MAC in the context of directional antennas becomes a chal-lenging issue. Unlike [1], our work aims to solve MAC protocol issues for using directional antennas in IEEE 802.16(d) mesh networks.
Differing from the previous work, our work does not enhance the CSMA/CA based MAC protocol. The IEEE 802.16(d) mesh network employing a novel dis-tributed election-based algorithm, and thus using directional antennas for such networks should use quite different techniques to solve the issues mentioned pre-viously. To best of our knowledge, our paper is the first work employing steerable directional antenna for the IEEE 802.16(d) mesh mode. We design an enhanced version of the distributed election algorithm to solve the contention of control message transmissions and the deafness problem, when adopting directional an-tennas in the IEEE 802.16(d) networks. To initialize such an IEEE 802.16(d) mesh directional network, we also propose a hash-function-based scheme to per-form neighboring node discovery and initial synchronization.
Chapter 3
Background
In this chapter, we first take an overview of the IEEE 802.16(d) mesh network. Second, the effect of the holdoff time value will be discussed. Lastly, we will introduce some works relative to using directional antennas in the IEEE 802.16 network.
3.1
Overview of IEEE 802.16(d) Mesh Networks
The IEEE 802.16(d) standard[5] defines an air interface of fixed broadband wireless access (BWA) system specifications providing high network throughput and low packet loss rate broadband communications. The BWA system uses the median based on the single-carrier modulation in the 10-66 GHz licensed bands or the orthogonal frequency division multiplexing (OFDM) in frequencies below 11 GHz. In the reference model of the standard, two protocol layers are defined. First, the median access control (MAC) layer comprises three sublayers, namely the service-specific convergence sublayer, the MAC common part sublayer (MAC CPS), and the security sublayer. Second, the physical layer defines multiple specifications for different frequency ranges and applications.
The MAC layer of the IEEE 802.16 network supports two operation modes for sharing wireless media. First, the point-to-multipoint (PMP) architecture is designed for one-hop communication between a base station (BS) and multiple
subscriber stations (SS). Second, the mesh topology mode is designed for a multi-hop wireless network in which any pair of one-multi-hop distancing SSs (including the BS) can communication with each other. In the mesh mode, the BS has a direct connection to backhaul services for SSs to communication with hosts outside the mesh network.
To avoid data transmission collisions, the IEEE 802.16 mesh mode provides two scheduling modes — the centralized and the distributed modes. The distributed mode is further divided into the coordinated and the uncoordinated scheduling, respectively. In this thesis, the proposed design is dedicated for the distributed coordinated scheduling mesh mode using directional antennas.
3.1.1
Node, Neighbor, Neighborhood, and Extended
Neigh-borhood
A node is a generic term for a BS and a SS in the mesh network. Stations with which a node can directly communicate are called the node’s one-hop neighbors or neighbors in brief. Neighbors of a node form a neighborhood and all the neighbors of the nodes in the neighborhood form a extended neighborhood. A node’s two-hop neighbors are the nodes in the extended neighborhood excluding one-two-hop neighbors.
3.1.2
Network Entry
In the mesh network, each SS is called a new node before finishes the network entry procedure. A new node cannot schedule data transmission until it finishes the network entry procedure and becomes a functional node. (The BS is a functional node when the mesh network starts.)
In the network entry procedure, a new node first listens to the mesh network configuration (MSH-NCFG) message in the air. While receiving MSH-NCFG mes-sages, the new node shall maintain a physical neighborhood list according to the information carried in the MSH-NCFG message. The new node then selects a
po-tential sponsoring node from its neighborhood and asks for opening a temporary sponsor channel by sending a mesh network entry (MSH-NENT) message to the sponsoring node. After the sponsoring node opens the sponsor channel, the new node can communication with its sponsoring node by the sponsor channel.
With the sponsor channel, the new node can start the registration procedure by transmitting a registration request (REG-REQ) message to the sponsoring node. When the sponsoring node received a REG-REQ message from the sponsor channel, it tunnels the message by prepending a UDP header and a IP header to the registration node, usually collocated with the BS. Then the registration node shall assign a unique mesh node ID in the same network for the new node and sends a registration response (REG-RSP) message to the sponsoring node. When sponsoring node receives the REG-RSP message from the registration node, it forwards the message to the new node. The new node completes the registration procedure as receiving the REG-RSP message, then it will ask the sponsoring node to close the sponsor channel and finishes the network entry procedure.
3.1.3
Contention Resolution for Transmitting Control
Mes-sages
In the distributed coordinated scheduling mode, a functional node periodically broadcasts MSH-NCFG or MSH-DSCH messages to its one-hop neighbors on the transmission opportunity won in the previous contention. A functional node carries its next transmission time and its one-hop neighbors’ next transmission time in the control message. When a node received a control message, it can update the neighborhood list, which contains one-hop and two-hop neighbors, with next transmission opportunities carried in the received control message.
In the standard, instead of a precise number, each transmission opportunity carried in the control messages is expressed by a 5-bit mx and a 3-bit exponent as follows:
2exponent· mx < transmission opportunity ≤ 2exponent· (mx + 1), where 0 ≤ mx ≤ 30, 0 ≤ exponent ≤ 7
(3.1)
A node can derive a transmission interval from mx and exponent unless the mx value is 31. In the case that the mx value is 31, the node shall consider the transmission interval of the corresponding neighbor is unknown.
In the mesh network, a node transmits control messages only on the trans-mission opportunity won in the mesh election algorithm, which is defined in the standard. The mesh election algorithm uses an eligible nodes list as an input to determine whether a node wins a certain transmission opportunity. The mesh election algorithm can ensures the resultant transmission opportunity is collision free within the extended neighborhood.
A functional node performs the mesh election algorithm to contend for a specific transmission opportunity. First, the functional node has to derive the eligible nodes list from its neighborhood list. If the contended transmission opportunity is in a neighbor’s transmission interval, the node should consider the neighbor as a eligible node and add the neighbor into the eligible nodes list.
Additionally, when a node determines the eligibility of a neighbor, it shall consider the neighbor’s holdoff time, which will be explained in Section 3.2.1. The neighbor’s holdoff time plus 2exponent · mx is the earliest subsequent transmission
opportunity of the neighbor. A node shall exclude the neighbor from the eligible nodes list when it contends for a transmission opportunity before the neighbor’s earliest subsequent transmission opportunity.
3.1.4
Distributed Coordinated Data Transmission
Schedul-ing
Three different types of the information element (IE) can be carried by the MSH-DSCH message for the distributed coordinated data transmission scheduling. Each
data transmission is established by exchanging these IEs between a requesting node and a granting node in a three-way handshake procedure. A node can start a data transmission after it completes the three-way handshake procedure. We explain these IEs as follows:
Request IE
It carries the amount of the requesting resource.
Availability IE
It indicates free mini-slot ranges in which the granting node can issue a grant.
Grants IE
It carries the information about a granted mini-slot range if the IE is sent by the granting node. If the IE is sent by the requesting node, it confirms a grant.
The three-way handshake procedure is performed between two nodes for es-tablishing a data transmission. The requesting node first transmits a MSH-DSCH message containing a request IE and one or more availability IEs. When receiving this message, the granting node finds a free mini-slot range, if exists, which is included in the mini-slot range indicated in the received availability IE. If a proper mini-slot range is found, the granting node grants this request by transmitting a grant IE to the requesting node. Lastly, the requesting node sends a MSH-DSCH message including the copy of the received grant IE to confirm the grant.
3.2
Dynamic Holdoff Time Setting
3.2.1
The Holdoff Time in the IEEE 802.16 Mesh Network
In the IEEE 802.16 mesh network, each node determines the next transmission opportunity by the mesh election algorithm introduced in Section 3.1.3. In this election algorithm, a network node cannot contend for the transmission opportu-nity immediately following the current one. The standard requires it to refrain
Time Current Txopp The 1st contending next txopp Holdoff Time
Won next txopp
Contention Time
Figure 3.1: The holdoff time before contending for next transmission opportunity.
from contending in a certain number of consecutive transmission opportunities, which is called the holdoff time as shown in Fig 3.1.
Holdoff Time = 2exponent+base,
where base = 4, 0 ≤ exponent ≤ 7
(3.2)
In the standard, the holdoff time value is defined as Equation (3.2). Although the exponent value can be variant in different networks, all nodes in the same network are required to be consistent in the holdoff time exponent value.
3.2.2
Dynamic Holdoff Time Approach
In [16], authors propose a two-phase holdoff time setting scheme that uses differ-ent holdoff time values in its network initialization phase and its data transmission phase. The proposed holdoff time setting scheme ensures success of network ini-tialization. Additionally, two versions of the holdoff time setting scheme in the data transmitting phase are also proposed in the article.
Static Holdoff Time Value Setting
In the static version defined in [16], each node is assigned a different holdoff time based on its two-hop neighborhood node number. A node with a dense extended
neighborhood is assigned a larger holdoff time. Contrarily, a node with sparse extended neighborhood is assigned a smaller holdoff time. The improved perfor-mance of this static holdoff time assignment approach is shown in the article.
Dynamic Holdoff Time Value Setting
A dynamic holdoff time assignment approach is proposed in [16]. In this approach, a node adjusts its holdoff time according to the bandwidth requirement in time. This approach can effectively decrease roundly half required time of the three-way handshake procedure used for distributed coordinated data scheduling in the mesh network.
Discussion on Holdoff Time Base
In [16], the effect of fixed holdoff time base required by the standard[5] is discussed. The authors propose how to change the holdoff time base without losing standard compliance and explain advantages of setting the holdoff time base to zero.
Chapter 4
Protocol Design
This chapter proposes a protocol design for the IEEE 802.16 mesh mode MAC layer. The protocol aims to provide a comprehensive suite of modifications for using directional antennas in the IEEE 802.16 mesh network. We first define problems of using directional antennas in the mesh network. We then go through the original implementation of standard[3] over the NCTUns platform[17]. In the following, we define some terminologies used in this chapter. Finally, we will describe modifications to each component of the MAC-layer module in detail.
4.1
Difficulty of using Directional-antenna in IEEE
802.16 Mesh Network
In the IEEE 802.16 mesh network, each network node should use the mesh election algorithm mentioned in Section 3.1.3 to determine its control message transmission timing. This algorithm requires a list of eligible contending nodes as an input to determine the wining node for each transmission opportunity. A network node should maintain its extended neighborhood (defined in Section 3.1.1) to derive the eligible contending node list for each transmission opportunity. The maintenance of a node’s extended neighborhood relies on periodically exchanging control messages among neighboring nodes. In an omni-directional-antenna network, the control
MAC-layer Module NCFG Scheduler DSCH Scheduler Network Entry Manipulator Physical Frame Manager
Figure 4.1: Components of the MAC-layer Module
messages transmitted by a network node can be received by its neighboring nodes due to the broadcast nature of wireless radio. But in a directional-antenna network, only neighbors in the coverage of the transmitting node’s antenna can receive its control message. The neighbors out of the coverage cannot update its maintained extended neighborhood from the information in the control message sent by the transmitting node. If network nodes do not exchange control messages with their neighbors in time, the contention resolution is likely to fail in a directional-antenna network. The failure of contention resolution will incur the collision of control messages and then the network cannot operate accurately. Even worse, the network cannot be successfully initialized at the beginning. In the following sections we describe how the IEEE 802.16 mesh network works well using directional antennas. Furthermore, it works more efficiently and capably under our design.
4.2
Introduction to the Omni-direction-antenna
MAC layer
To clearly explain our modifications to the omni-directional-antenna MAC-layer implementation, we first introduce the essential components of the MAC-layer module in this section.
As shown in Fig 4.1, the MAC-layer module is mainly divided into four compo-nents — NCFG scheduler, DSCH scheduler, network entry manager, and physical frame manager. In the following, we explain how these four components are mod-ified to realize a directional-antenna version of the MAC-layer module.
time ... Network Control Subframe Data Subframe Schedule Control Subframe FRAMSN: 0 NENTSN: 1 NCFGSN: [1,NNCFG] DSCHSN: 0 FRAMSN: 1 NENTSN: 1 NCFGSN: [1,NNCFG] DSCHSN: [1,NDSCH] FRAMSN: 2 NENTSN: 1 NCFGSN: [1,NNCFG] DSCHSN: [NDSCH + 1,2∙NDSCH] FRAMSN: K NENTSN: 1 NCFGSN: [1,NNCFG] DSCHSN: [(K – 1)∙NDSCH + 1,K∙NDSCH] FRAMSN: K + 1 NENTSN: 2 NCFGSN: [NNCFG + 1,2∙NNCFG] DSCHSN: [(K – 1)∙NDSCH + 1,K∙NDSCH] K frames FRAMSN: K + 2 NENTSN: 2 NCFGSN: [NNCFG + 1,2∙NNCFG] DSCHSN: [K∙NDSCH + 1,(K + 1)∙NDSCH] Network Entry Network Config ... Network Config NNCFG Central Scheduling Central Config ... Distributed Scheduling NDSCH Distributed Scheduling Distributed Scheduling Network Config Network Config
Figure 4.2: Mesh Frame Structure.
Physical Frame Manager
For keeping track of the use of sophisticated mesh mode physical frame, we use an individual component for maintaining physical frame operating states and a variety of sequence numbers (i.e., frame counters).
Fig 4.2 shows the IEEE 802.16 mesh mode frame structure defined in [5]. K is the number of schedule control subframes between two network control subframes. Each network control subframe consists of one network entry and NN CF G network
configuration transmission opportunities for MSH-NENT and MSH-NCFG mes-sages sending, respectively. In the same fashion, NDSCH distributed scheduling
transmission opportunities are used for sending MSH-DSCH messages. Follow-ing the control subframe, the data subframe is divided into mini-slots for data transmission.
Four sequence numbers are maintained in this component:
FRAMSN
Frame sequence number, increased by one for every frame.
NENTSN
Network entry transmission opportunity sequence number, increased by one for every network control subframe.
Network configuration transmission opportunity sequence number, increased by NN CF G for every network control subframe.
DSCHSN
Distributed scheduling transmission opportunity sequence number, increased by NDSCH for every schedule control subframe.
The MAC-layer module uses these sequence numbers to determine which type of control messages or data should be sent in the current frame.
Network Entry Process Manager
The network entry process manager performs a network-attaching procedures when a new node is attaching itself to the network. Besides, for a functional node its network entry process manager will perform sponsoring procedures to help new nodes attach themselves to the network.
Control Message Schedulers
NCFG and DSCH schedulers are used for scheduling transmissions of MSH-NCFG and MSH-DSCH messages, respectively. A control message schedule determines the next transmission opportunity of the control messages using the mesh election algorithm explained in Section 3.1.3. To this end, it maintains the latest next transmission opportunity and holdoff exponent value of neighboring nodes. When receiving a MSH-NCFG or MSH-DSCH message from a neighboring functional node, the control message scheduler updates the above information which will be used by the mesh election algorithm.
Data Scheduler
The data scheduler manages mini-slot allocation in the data-plane. It maintains the status of each mini-slot allocation, which is defined as a range of mini-slots spanning a certain number of frames. The status of a mini-slot allocation indicates if an allocation is ready to transmit/receive data packets. The data scheduler thus
Index:0 Index:3
Index:1
Index:2
(a) Antenna beam width π2.
Index:0
Index:3
Index:1
Index:2
Index:4 Index:5
(b) Antenna beam width π3.
Figure 4.3: Antenna domain definition.
can determine whether a request to a mini-slot allocation can be accepted using the maintained allocation statuses. The scheduler will try to find an available range of mini-slots for the MAC-layer module when it has data to send. Besides, the data scheduler has to record mini-slot allocations used by neighboring nodes to avoid conflicting the schedules of neighboring nodes.
4.3
Directional-antenna Related Terminologies
We define an antenna domain as an area covered by a sector antenna. Each domain is given a unique number, called “domain index ”. Fig 4.3 illustrates a node’s antenna domains from the geometric view. The rule for indexing such domains, which use antennas with a beam width of radian B, is as follows. A domain i covers angles between B · i ±12 mod 2π ∀ 0 ≤ i ≤ 2πB. As Fig 4.3 (b) shows, the four domains are B · i ± 12 mod 2π ∀ 0 ≤ i ≤ 3
4.4
Modified Network Entry Process
4.4.1
Issues involved in Using Directional-antenna
In an omni-directional-antenna network, during the initialization stage a new node learns the existence of its neighboring functional nodes by monitoring their MSH-NCFG messages. In a directional-antenna network, however, two undesired issues may arise. One is that a new node cannot start the network entry procedure if it cannot detect any neighboring functional nodes. In a directional-antenna network, a new node cannot predict when and in which direction its neighboring nodes may transmit their MSH-NCFG messages before successfully attaching to this network. The other issue is that even if a new node has completed the synchronization and selected a proper sponsoring node, it cannot determine when to send its MSH-NENT message because it cannot know when its sponsoring node is ready to receive this message (i.e., point its antenna to cover this new node). Similarly, a functional node cannot predict when or from which direction it can receive a new node’s MSH-NENT message. In the following, we explain how to solve these two problems.
4.4.2
Node Selection for Synchronization and Network
En-try
As mentioned previously, a new node cannot reliably receive MSH-NCFG messages in a directional antenna network. To solve this problem, a functional node and a new node should determine when and where their antennas must point to each other. To solve this problem, we define the following hash function:
Is tmp txopp equal to one of hash txopps?
next txopp =
the result of the mesh election algorithm
For each non-functional neighbor in current antenna domain, calculate hash txopps by
He(My Node ID, Peer Node ID)
No
Yes
Return next txopp
Yes No Is there any non-functional neighbor? Yes Return tmp txopp tmp txopp = current txopp + 1
No
tmp txopp = tmp txopp + 1 Is tmp txopp equal to next txopp?
Figure 4.4: The refined mesh election algorithm.
He(nT x, nRx) = ((nT x & Nm) Nbit) | (nRx & Nm),
where nT x is transmitting node ID
nRx is receiving node ID
Nm is a mask value that may be adjusted under different networks.
Nbit is the number of bits of Nm
(4.1)
The resulted hash value is the NCFGSN (defined in Section 4.2) for a func-tional node to transmit its MSH-NCFG message. The value of Nm may limit the
scalability of network. We fix the value of Nm to be ‘0x1f’ in our simulations.
We then refine the original mesh election algorithm with Equation (4.1), which is shown in Fig 4.4. Each functional node uses this refined mesh election algorithm
to determine its next transmission opportunity.
Similarly, a new node can use the hash function to know when its neighboring function node will transmit MSH-NCFG messages to itself. Different from the functional node, the new node just uses the hash function to predict the trans-mission timing of its neighboring functional nodes’ control messages, instead of deriving its control message transmission timing.
4.4.3
Antenna Orientation for MSH-NENT message
Trans-mission
In a directional-antenna network, antenna orientation is key to the success of network operation. For a pair of a sending and receiving nodes, they should point their antennas to each other at the same time to correctly transmit/receive their control messages. Moreover, a sponsoring node and a new node should know the transmission timing of their control messages. Otherwise, the new node cannot proceed its network entry process. To achieve this, we design two hash functions, one of which is for sponsoring nodes and the other is for new nodes.
Hf(t) = t mod N, where N is the number of domains (4.2)
where t is the current NENTSN of a functional node
Hn(t) = BH + π 0 ≤ BH < π BH − π π ≤ BH < 2π (4.3) where BH = B · Hf(t)
B is the antenna beam width
t is the current NENTSN of a new node
After a new node synchronizes with a neighboring functional node, the NENTSN will be consistent with the network. (i.e., the NENTSN of the new node will be the same as all functional nodes in this network) In the beginning, all functional
Functional Node New Node New Node (a) Case N = 4, t = kN Functional Node New Node New Node (b) Case N = 4, t = kN + 2
Figure 4.5: Antennae orientation for MSH-NENT transmission.
nodes use Equation (4.2) to derive the MSH-NENT message transmission timing of new nodes. Similarly, new nodes use Equation (4.3) to know the orientation of their antennas at any time.
With such a design, each new node can exchange control messages with its sponsor node every N MSH-NENT transmission opportunities and thus proceed its network entry process. Fig 4.5 (a) shows a cases with N = 4, t = kN and Fig 4.5 (b) shows a case with kN + 2 where k ∈ N.
4.5
Modified Contention Resolution for Control
Messages
4.5.1
Issues of using Directional-antenna
Functional nodes in an IEEE 802.16 mesh network requires periodically trans-mitting MSH-NCFG and MSH-DSCH messages to exchange their collected local information for network maintenance. For example, these two messages contain
the holdoff time exponent values and the next transmission opportunities of nodes adjacent to the transmitting node. On receiving such a message, the receiving node should use the received information to update its local neighborhood list. As a functional node is going to transmit a MSH-NCFG or a MSH-DSCH message, it will perform the mesh election algorithm with this updated local neighborhood list as input to determine its next transmission opportunity.
In a network employing directional antennas, however, exchanges of MSH-NCFG and MSH-DSCH messages are possible only when two neighboring nodes’ antenna beams can cover each other. In such a condition, updating a node’s local neighborhood list is more challenging and difficult. For example, in Fig 4.6 only node B can successfully receive the message transmitted by the sending node. In such a condition, nodes A and C cannot receive the latest next transmission opportunity and holdoff time exponent value of the sending node. This is very likely to result in failed synchronization between the sending node and these two nodes. Even worse, several nodes in the sending node’s extended neighborhood (defined in Section 3.1.1) may not obtain the latest scheduling information of the sending node if they update the sending node’s scheduling information based on only nodes A and C’s control messages.
4.5.2
Antenna Orientation for MSH-NCFG and MSH-DSCH
messages Transmission
Consider node C in Fig 4.6. If node C is able to know when the sending node will send its control message, node C can point its antenna to the sending node in time. To achieve this, the sending node has to notify other nodes of its next transmission opportunity precisely.
However, in [5], instead of a precise number, a node’s next transmission op-portunity is represented by an interval of length 2exp (explained in Section 3.1.3). Therefore, a node cannot know its neighboring nodes’ next transmission opportuni-ties accurately. Besides, more than one nodes may transmit their control messages
A B
C
Sending Node
Figure 4.6: Receiving control messages from the sending node is more difficult in directional-antenna networks.
during the same interval. In such a condition, the receiving node should point its antenna to only one of these nodes during the interval and thus will miss control messages transmitted by the other neighboring nodes.
Due to the limit of the current standard, it is impossible to describe the accurate transmission opportunity of a network node using the currently-defined control message format. To solve this problem, we extend the use of the formats of MSH-NCFG and MSH-DSCH messages to carry additional offset information for each node’s next transmission interval. Using such a design, a receiving node can obtain the starting transmission opportunity of a transmitting node’s next transmission interval and an offset value. As shown in Fig 4.7, the receiving node can then derive the accurate next transmission opportunity of the transmitting node by adding the starting transmission opportunity to the offset value.
While the neighbors’ next transmission opportunities are announced exactly, what the receiving node has to do is to turn its antenna to the node using the current transmission opportunity for control message sending.
Time
Mx∙2exp
Exactly Next Txopp Next Txopp Interval
Offset
(Mx + 1)∙2exp
Figure 4.7: The offset used for expressing the transmission opportunity exactly.
4.5.3
Modifications of Control Message Scheduler
To realize the design described in the previous section, some modifications are applied to the control message scheduler in the following sections.
Multiple Next Transmission Opportunities (MNTO) Maintenance
Due to the nature of the directional antenna, it is impossible for a node to broad-cast its next transmission opportunity to the nodes in all antenna domains in a transmission opportunity. Nevertheless, the node can tell all nodes in just one antenna domain its next transmission opportunity when its antenna beam covers the whole targeted domain.
For each antenna domain of the transmission node, we maintain an individual next transmission opportunity in the control message scheduler. In other words, the scheduler should maintain N different next transmission opportunities simul-taneously in its internal data structure, where N denotes the number of antenna domains of the node.
Directional-antenna version Mesh Election Algorithm (DMEA)
In MNTO scheme, all antenna domains’ next transmission opportunities must be pairwisely different. If the control message scheduler uses the original version mesh election algorithm, it may get a next transmission opportunity which was chosen for another antenna domain. Since the mesh election algorithm is a deterministic algorithm, the same result would be yielded if the input of the algorithm is not
Yes i = i + 1 txoppS = i-th small one txopp
in txopps for all antenna domains
No Set the contention start txopp = txoppS
txoppS = current txopp
i = 1
return (tmp txopp,tmp exp) Get (tmp txopp,tmp exp) by the DEDA. If operation fails, set the error flag.
Does the error flag bave been set?
Figure 4.8: Directional-antenna version Mesh Election Algorithm.
changed. To coordinate with the MNTO, we refine the mesh election algorithm as shown in Fig 4.8. We will explain this algorithm more elaborately in the following paragraphs.
Recall in Section 3.1.3, the algorithm by default uses the current transmission opportunity plus holdoff time as the first contending transmission opportunity, which is illustrated in Fig 3.1. If the control message scheduler finds the resul-tant next transmission opportunity is used by another antenna domain, it changes the first contending transmission opportunity to the second smallest transmission opportunity in all antenna domains. If a duplicated transmission opportunity is obtained, it pushes the first contending point to the third smallest one. In any case, a distinct transmission opportunity will be chosen when using the largest trans-mission opportunity as the first contending point. Fig 4.9 is helpful to understand this idea.
After the first contending transmission opportunity is settled, we use DEDA to get the next transmission opportunity and holdoff time exponent value. (DEDA will soon be explained in Section 4.5.3) This information are told to the nodes in the current antenna domain. Hence, these nodes can determine when they should turn their antennas back to the sending node again.
3rd round of Mesh Election Algorithm 2nd round of Mesh Election Algorithm 1st round of Mesh Election Algorithm Time
Next Txopp used by the domain 3
Next Txopp used by the domain 1
Next Txopp used by the domain 0 Next Txopp won
in 1st round
Next Txopp won in 2nd round
Next Txopp won In 3rd round
Holdoff Time
Current Txopp
Figure 4.9: Determine the next transmission opportunity for the antenna domain with index 2
In summary, DMEA is used by the control message scheduler before sending a message to one of its antenna domains. The schedule uses DMEA to determine the next transmission opportunity and holdoff time exponent value for the domain in the current antenna direction. This information will be carried to all the receiving node in the domain by embedding it into the control message. Besides, it will be used for updating the next transmission opportunity of this domain in the scheduler’s internal data structure.
Dynamic Holdoff Time Exponent Determination Algorithm
For the sake of the network performance, the directional-antenna network op-erating is based on the static holdoff time assignment version mentioned in the Section 3.2.2. In such a version, each network node is assigned a fixed holdoff time exponent value individually depending on its neighborhood size. Thus, a network node uses this holdoff time value in DMEA to obtain different next transmission opportunities for all antenna domains.
So far, the directional-antenna based mesh network works well but will not work efficiently under the present design. Compared with the omni-directional-antenna network, a node has to use roundly N times control messages for notifying nodes in all the antenna domain to update the next transmission opportunity of the node (N is the number of antenna domains). The transmitting interval between two control messages for each neighboring node is extended about N times since the
control messages are sent to each antenna domain in a round-robin like fashion. This phenomenon increases the delay of three-way handshake using MSH-DSCH messages.
The reason for a lower frequency of sending control message is that a consis-tent holdoff time is used for all antenna domains of a node. When the holdoff time used in DMEA for determining the current antenna domain’s next transmission opportunity is the same as that used for the previous one, it is very likely to obtain the same transmission opportunity by using DMEA with unchanged holdoff time value. Especially, if the neighborhood list has not been updated before determining the next transmission opportunity of another antenna domain, the same transmis-sion opportunity would be always obtained. This is because the neighborhood list is the primary input to the mesh election algorithm (Recall Section 3.1.3).
Fig 4.10 (a) explains this problem more clearly if a static holdoff time value is used in DMEA. A node first tries to get the next transmission opportunity for the antenna domain with index 2. The first round of DMEA fails because it chooses a transmission opportunity used by the antenna domain with index 1. Fortunately, a free transmission opportunity is generated in the second round of DMEA. Sequentially, the node uses two iterations of DMEA to determine the next transmission opportunity for the antenna domain with index 1 since the first chosen transmission opportunity just meets the one which is determined previously for the antenna domain with index 2. In this case, we only consider two antenna domains for simplicity. Actually, it would be worse when more antenna domains exist.
To decrease the probability of resulting the same transmission opportunity using DMEA for different antenna domains, we introduce a dynamic holdoff time scheme. In each DMEA round, instead of using only the preassigned holdoff time exponent value, DMEA changes the holdoff time exponent value if it cannot find a free transmission opportunity by using the previous one. Fig 4.10 (b) shows this idea more clearly. When DMEA detects that a free transmission opportunity cannot be found by using a holdoff time of 16, it changes the holdoff time to 8
2nd round of Mesh Election Algorithm for domain 1 1st round of Mesh Election Algorithm for domain 1 2nd round of Mesh Election Algorithm for domain 2 1st round of Mesh Election Algorithm for domain 2
Current Txopp used by the domain 1 Txopp won in 1st round Txopp won in 2nd round Holdoff Time Time 4 8 12 16 20 24 28 32 36 40 44 48 52 56 60 64 Txopp won in 1st round Txopp won in 2nd round Holdoff Time
Next Txopp used by the domain 1 Current Txopp used
by the domain 2 0
Next Txopp used by the domain 2
(a) Two antenna domains contend the next transmission opportunity using static holdoff time in DMEA 1st round of Mesh Election Algorithm for domain 1 1st round of Mesh Election Algorithm for domain 2
Next Txopp used by the domain 1 Txopp won in 1st round Holdoff Time Time 4 8 12 16 20 24 28 32 36 40 44 48 52 56 60 64 Txopp won in 1st round Holdoff Time
Current Txopp used by the domain 2
0
Next Txopp used by the domain 2
Current Txopp used by the domain 1
(b) Two antenna domains contend the next transmission opportunity using dynamic holdoff time in DMEA
Figure 4.10: Using different holdoff time values for different antenna domains.
and runs the mesh election algorithm again. In this case, a valid transmission opportunity for the antenna domain with index 2 is found in the first round of DMEA. Moreover, the transmission opportunity is found in the first round of DMEA without changing of the holdoff time.
The algorithm how DMEA changes the holdoff time is called DEDA, which is presented in Fig 4.11. Some details of DEDA are explained below.
First, we define two constants as follows:
β The maximum amount of increased exponent.
The constant α is used for restricting the number of exponent that can be decreased while β is used for restricting the number of exponent that can be increased in DEDA. In other word, if DEDA starts with exponent e, the exponent used by DEDA will not excess e + β or less than e − α. These two constants should be consistent for all network nodes using DEDA.
Initially, DEDA uses the preassigned holdoff time exponent value explained in Section 3.2.2 as the exponent value when it turns from a new node into a functional node. Subsequently, it uses the previous adopted exponent value of the current antenna domain as the initial exponent value. When sending regular control message, DEDA first adjusts the exponent value increasingly until the exponent value exceeds the defined threshold. If DEDA cannot find an appropriate exponent by increasing the exponent value, it resets the exponent value and adjusts the value decreasingly. DEDA may fail if it is unable to get an proper exponent value which can help the mesh election algorithm to get a transmission opportunity free from overlapping others antenna domains’ transmission opportunities. In such a condition, the control message scheduler will resort to DMEA to start the next round.
Exploiting the Information of Three-way Handshake Procedure
Recall that the three-way handshake procedure used for establishing a data sched-ule requires transmitting three MSH-DSCH messages (Section 4.2). Without con-sidering the dynamic bandwidth needs of nodes, DEDA may choose a large ex-ponent for a node when determining the next transmission opportunity of MSH-DSCH regardless whether the node has data to send. Thus, the delay between two consecutive MSH-DSCH messages for the three-way handshake procedure can be large. This will result in decreased per-hop (as well as end-to-end) data trans-mission delays and increased per-hop (as well as end-to-end) data transtrans-mission throughputs.
