VIEWS: 6 PAGES: 74 POSTED ON: 11/1/2011 Public Domain
Quorum-Based Asynchronous Power-Saving Protocols for IEEE 802.11 Ad Hoc Networks Presented by Jehn-Ruey Jiang Department of Computer Science and Information Engineering National Central University To Rest, to Go Far! Outline IEEE 802.11 Ad hoc Network Power Saving Problem Asynchronous Quorum-based PS Protocols Optimal Asyn. Quorum-Based PS Protocols Analysis and Simulation Conclusion Outline IEEE 802.11 Ad hoc Network Power Saving Problem Asynchronous Quorum-based PS Protocols Optimal Asyn. Quorum-Based PS Protocols Analysis and Simulation Conclusion IEEE 802.11 Overview Approved by IEEE in 1997 Extensions approved in 1999 Standard for Wireless Local Area Networks ( WLAN ) IEEE 802.11 Family(1/2) 802.11a: 6 to 54 Mbps in the 5 GHz band 802.11b (WiFi, Wireless Fidelity): 5.5 and 11 Mbps in the 2.4 GHz band 802.11g: 54 Mbps in the 2.4 GHz band IEEE 802.11 Family(2/2) 802.11c: support for 802.11 frames 802.11d: new support for 802.11 frames 802.11e: QoS enhancement in MAC 802.11f: Inter Access Point Protocol 802.11h: channel selection and power control 802.11i: security enhancement in MAC 802.11j: 5 GHz globalization IEEE 802.11 Market Source: Cahners In-Stat ($ Million) Infrastructure vs Ad-hoc Modes infrastructure network AP AP wired network AP Multi-hop ad hoc network ad-hoc network ad-hoc network Ad hoc Network Applications Battlefields Disaster rescue Spontaneous meetings Outdoor activities Outline IEEE 802.11 Ad hoc Network Power Saving Problem Asynchronous Quorum-based PS Protocols Optimal Asyn. Quorum-Based PS Protocols Analysis and Simulation Conclusion Power Saving Battery is a limited resource for portable devices Battery technology does not progress fast enough Power saving becomes a critical issue in MANETs, in which devices are all supported by batteries Solutions to Power Saving PHY Layer: transmission power control Huang (ICCCN’01), Ramanathan (INFOCOM’00) MAC Layer: power mode management Tseng (INFOCOM’02), Chiasserini (WCNC’00) Network Layer: power-aware routing Singh (ICMCN’98), Ryu (ICC’00) Transmission Power Control Tuning transmission energy for higher channel reuse Example: A is sending to B (based on IEEE 802.11) Can (C, D) and (E, F) join? No! Yes! B C D A F E Power Mode Management doze mode vs. active mode Example: A is sending to B Does C need to stay awake? No! It can turn off its radio B to save energy! A But it should turn on its C radio periodiclally for possible data comm. Power-Aware Routing Routing in an ad hoc network with energy- saving (prolonging network lifetime) in mind Example: N1 + N2 + SRC – – DES T + Better!! + – – + + N3 N4 – – Our Focus Among the three solutions: PHY Layer: transmission power control MAC Layer: power mode management Network Layer: power-aware routing IEEE 802.11 PS Mode(2/2) Environments: Infrastructure (O) Ad hoc (infrastructureless) Single-hop (O) Multi-hop IEEE 802.11 PS Mode(1/2) An IEEE 802.11 Card is allowed to turn off its radio to be in the PS mode to save energy Power Consumption: (ORiNOCO IEEE 802.11b PC Gold Card) Vcc:5V, Speed:11Mbps PS for 1-hop Ad hoc Networks (1/3) Time axis is divided into equal-length intervals called beacon intervals In the beginning of a beacon interval, there is ATIM window, in which hosts should wake up and contend to send a beacon frame with the backoff mechanism for synchronizing clocks Beacon Interval Beacon Interval Beacon Interval Beacon Interval ATIM Power Window Saving Mode Host Beacon PS for 1-hop Ad hoc Networks (2/3) A possible sender also sends ATIM (Ad hoc Traffic Indication Map) message with DCF procedure in the ATIM window to its intended receivers in the PS mode ATIM demands an ACK. And the pair of hosts receiving ATIM and ATIM-ACK should keep themselves awake for transmitting and receiving data PS for 1-hop Ad hoc Networks (3/3) Target Beacon Transmission Time (TBTT) Beacon Interval Beacon Interval ATIM power ATIM active state Window saving Window mode Host A ATIM Beacon No ATIM means data BTA=2, BTB=5 no data to send frame or to receive ATIM power ATIM Window saving Window mode Host B Beacon ACK ACK PS: m-hop Ad hoc Network Problems: Clock Synchronization it is hard due to communication delays and mobility Network Partition unsynchronized hosts with different wakeup times may not recognize each other Clock Drift Example Max. clock drift for IEEE 802.11 TSF (200 DSSS nodes, 11Mbps, aBP=0.1s) Network-Partitioning Example C D ╳ F The red ones do not The blue ones do not A Network know the existence of Partition the red ones, not to the blueones, not to ╳ mention the time when B E they are awake. Host A ATIM Host B ╳ window Host C Host D ╳ Host E Host F Asynchronous PS Protocols (1/2) Try to solve the network partitioning problem to achieve Neighbor discovery Wakeup prediction without synchronizing hosts’ clocks Asynchronous PS Protocols (2/2) Three asyn. PS protocols by Tseng: Dominating-Awake-Interval Periodical-Fully-Awake-Interval Quorum-Based Ref: “Power-Saving Protocols for IEEE 802.11-Based Multi-Hop Ad Hoc Networks,” Yu-Chee Tseng, Chih-Shun Hsu and Ten-Yueng Hsieh InfoCom’2002 Outline IEEE 802.11 Ad hoc Network Power Saving Problem Asynchronous Quorum-based PS Protocols Optimal Asyn. Quorum-Based PS Protocols Analysis and Simulation Conclusion Numbering beacon intervals 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 2 … Beacon interval n consecutive beacon intervals are numbered as 0 to n-1 0 1 2 3 4 5 6 7 And they are organized 8 9 10 11 as a n n array 12 13 14 15 Quorum Intervals (1/4) Intervals from one row and one column are called quorum intervals 0 1 2 3 Example: Quorum intervals are 4 5 6 7 numbered by 8 9 10 11 2, 6, 8, 9, 10, 11, 14 12 13 14 15 Quorum Intervals (2/4) Intervals from one row and one column are called quorum intervals 0 1 2 3 Example: Quorum intervals are 4 5 6 7 numbered by 8 9 10 11 0, 1, 2, 3, 5, 9, 13 12 13 14 15 Quorum Intervals (3/4) Any two sets of quorum intervals have two common members For example: The set of quorum intervals {0, 1, 2, 3, 5, 9, 13} and 0 1 2 3 the set of quorum intervals 4 5 6 7 {2, 6, 8, 9, 10, 11, 14} have 8 9 10 11 two common members: 12 13 14 15 2 and 9 Quorum Intervals (4/4) Host D 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Host C 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 2 overlapping quorum intervals Host D 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Host C 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Even when the beacon interval numbers are not aligned (they are rotated), there are always at least two overlapping quorum intervals Structure of quorum intervals Networks Merge Properly C D F A B E Host A ATIM Host B window Host C Beacon window Monitor window Host D Host E Host F Short Summary There is an asynchronous power- saving protocol that achieves asynchronous neighbor discovery Hearing beacons twice or more in every n consecutive beacon intervals wakeup prediction via a simple quorum concept. Observation 1 It is a simple grid quorum system [Maekawa 1985] in Tseng’s work. There are many more complicated quorum systems in the literature of distributed system: FPP [Maekawa 1985], Tree [Agrawal 1990], Hierarchical[Kumar 1991], Cohorts [Jiang 1997], Cyclic [Luk 1997], Torus [Lang 1998], etc. Question: Can these quorum systems be directly applied to solve the power-saving problem in a MANET? The Answer Is … Not all quorum systems can be used here! Counter example: { {1}} under {1,2,3} Onlythose quorum systems with the rotation closure property can be used! Observation 2 Smaller quorums are better because they imply lower active ratio (better energy-efficiency) But quorums cannot be too small less the quorum system does not satisfy the rotation closure property Question 1: What is the smallest quorum size? Question 2: Is there any quorum systems to have the smallest quorum size? Outline IEEE 802.11 Ad hoc Network Power Saving Problem Asynchronous Quorum-based PS Protocols Optimal Asyn. Quorum-Based PS Protocols Analysis and Simulation Conclusion What are quorum systems? Quorum system: a collection of mutually intersecting subsets of a universal set U, where each subset is called a quorum E.G. {{1, 2},{2, 3},{1,3}} is a quorum system under U={1,2,3} A quorum system is a collection of sets satisfying the intersection property Rotation Closure Property (1/3) Definition. Given a non-negative integer i and a quorum H in a quorum system Q under U = {0,…, n1}, we define rotate(H, i) = {j+ijH} (mod n). E.G. Let H={0,3} be a subset of U={0,…,3}. We have rotate(H, 0)={0, 3}, rotate(H, 1)={1,0}, rotate(H, 2)={2, 1}, rotate(H, 3)={3, 2} Rotation Closure Property (2/3) Definition. A quorum system Q under U = {0,…, n1} is said to have the rotation closure property if G,H Q, i {0,…, n1}: G rotate(H, i) . Rotation Closure Property (3/3) For example, Q1={{0,1},{0,2},{1,2}} under U={0,1,2} Q2={{0,1},{0,2},{0,3},{1,2,3}} under U={0,1,2,3} Because {0,1} rotate({0,3},3) = {0,1} {3, 2} = Closure Examples of quorum systems Majorityquorum system Tree quorum system Hierarchical quorum system Cohorts quorum system ……… Optimal Quorum System (1/2) Quorum Size Lower Bound for quorum systems satisfying the rotation closure property: k, where k(k-1)+1=n, the cardinality of the universal set, and k-1 is a prime power (k n ) Optimal Quorum System (2/2) Optimal quorum system FPP quorum system Near optimal quorum systems Grid quorum system Torus quorum system Cyclic (difference set) quorum system Outline IEEE 802.11 Ad hoc Network Power Saving Problem Asynchronous Quorum-based PS Protocols Optimal Asyn. Quorum-Based PS Protocols Analysis and Simulation Conclusion Analysis (1/3) Active Ratio: the number of quorum intervals over n, where n is cardinality of the universal set Neighbor Sensibility (NS) the worst-case delay for a PS host to detect the existence of a newly approaching PS host in its neighborhood Analysis (2/3) Analysis (3/3) Optimal! Simulation Model Area: 1000m x 1000m Speed: 2Mbps Radio radius: 250m Battery energy: 100J. Traffic load: Poisson Dist. , 1~4 routes/s, each having ten 1k packets Mobility: way-point model (pause time: 20s) Routing protocol: AODV Simulation Parameters Unicast send 454+1.9 * L Broadcast send 266+1.9 * L Unicast receive 356+0.5 * L Broadcast receive 56+0.5 * L Idle 843 Doze 27 L: packet length Unicast packet size 1024 bytes Broadcast packet size 32 bytes Beacon window size 4ms MTIM window size 16ms Simulation Metrics ratio Survival Neighbor discovery time Throughput Aggregate throughput Simulation Results (1/10) E-torus quorum system Cyclic quorum system Always Active Survival ratio vs. mobility (beacon interval = 100 ms, 100 hosts, traffic load = 1 route/sec). Simulation Results (2/10) Neighbor discovery time (ms) 3000 2500 A faster host can be 2000 discovered in 1500 shorter time. 1000 C(98) 500 E(7x14) 0 0 5 10 15 20 pe d e Moving s e (m/s c) Neighbor discovery time vs. mobility (beacon interval =100 ms, 100 hosts, traffic load = 1 route/sec). Simulation Results (3/10) For the throughput: AA>E(7x74)>C(98) For the aggregate throughput: C(98)>E(7x74)>AA Throughput vs. mobility (beacon interval = 100 ms, 100 hosts, traffic load = 1 route/sec). Simulation Results (4/10) Survival ratio vs. beacon interval length (100 hosts, traffic load = 1 route/sec, moving speed = 0~20 m/sec with mean = 10m/sec). Simulation Results (5/10) 16000 Neighbor discovery time (ms) 14000 C(98) 12000 E(7x14) 10000 8000 6000 4000 2000 0 100 200 300 400 Beacon interval (ms) Neighbor discovery time vs. beacon interval length (100 hosts, traffic load = 1 route/sec, moving speed = 0~20 m/sec with mean = 10m/sec). Simulation Results (6/10) Throughput vs. beacon interval length (100 hosts, traffic load = 1 route/sec, moving speed = 0~20 m/sec with mean =10m/sec). Simulation Results (7/10) Survival ratio vs. traffic load (beacon interval = 100 ms, 100 hosts, mobility = 0~20 m/sec with mean = 10 m/sec). Simulation Results (8/10) Throughput vs. traffic load (beacon interval =100 ms, 100 hosts, mobility = 0~20 m/sec with mean = 10 m/sec). Simulation Results (9/10) Survival ratio vs. host density (beacon interval = 100ms, traffic load 1 route/sec, mobility = 0~20 m/sec with mean= 10 m/sec). Simulation Results (10/10) Throughput vs. host density (beacon interval = 100ms, traffic load 1 route/sec, mobility = 0~20m/sec with mean= 10 m/sec). Outline IEEE 802.11 Ad hoc Network Power Saving Problem Asynchronous Quorum-based PS Protocols Optimal Asyn. Quorum-Based PS Protocols Analysis and Simulation Conclusion Conclusion Quorum systems with the rotation closure property can be translated to an asyn. PS protocol. The active ratio is bounded by 1/ n, where n is the number of a group of consecutive beacon intervals. Optimal, near optimal and adaptive AQPS protocols save a lot of energy w/o degrading performance significantly Publication ICPP’03 Best Paper Award ACM Journal on Mobile Networks and Applications Future work To incorporate the clustering concept into the design of hybrid (syn. and asyn.) power saving protocols (NSC 93-2213-E-008-046-) To design more flexible adaptive asyn. power saving protocols with the aid of the expectation quorum system (a novel quorum system which is a general form of probabilistic quorum systems) (93CAISER-中央大學分部計畫) To incorporate power saving mode management to wireless sensor networks with comm. and sensing coverage in mind (中大新進教師學術研究 經費補助計畫) Thanks! FPP quorum system Proposed by Maekawa in 1985 For solving distributed mutual exclusion Constructed with a hypergraph An edge can connect more than 2 vertices FPP:Finite Projective Plane A hypergraph with each pair of edges having exactly one common vertex Also a Singer difference set quorum system FPP quorum system Example 5 A FPP quorum system: { {0,1,2}, 3 4 {1,5,6}, 6 {2,3,6}, {0,4,6}, 0 1 2 {1,3,4}, {2,4,5}, {0,3,5} } Torus quorum system 0 1 2 3 4 5 { {1,7,13,8,3,10}, 6 7 8 9 10 11 {5,11,17,12,1,14},…} 12 13 14 15 16 17 One half column cover in a wrap around manner One full column For a tw torus, a quorum contains all elements from some column c, plus w/2 elements, each of which comes from column c+i, i=1.. w/2 Cyclic (difference set) quorum system Def: A subset D={d1,…,dk} of Zn is called a difference set if for every e0 (mod n), there exist elements di and djD such that di-dj=e. {0,1,2,4} is a difference set under Z8 { {0, 1, 2, 4}, {1, 2, 3, 5}, {2, 3, 4, 6}, {3, 4, 5, 7}, {4, 5, 6, 0}, {5, 6, 7, 1}, {6, 7, 0, 2}, {7, 0, 1, 3} } is a cyclic (difference set) quorum system C(8) E-Torus quorum system Trunk E(t x w, k) Branch Branch cyclic Branch Branch cyclic