Analysis and Design of Link Metrics for Quality Routing in Wireless Multi-hop Networks

Transcription

1 Analysis and Design of Link Metrics for Quality Routing PhD Thesis Defense by Nadeem JAVAID Dec 15, 2010 Thesis Director Prof. Karim DJOUANI Jury : Rapporteur B.J. VAN WYK Prof. Tshwane University of Technology - South Africa Rapporteur J. BEN-OTHMAN MDC (HDR), Université de Versailles - France Examinateur K. BARKAOUI Prof. CNAM, Paris - France Examinateur O. BRUN Directeur de recherche, LAAS-CNRS, Toulouse - France Dir. de thèse K. DJOUANI Prof. Université Paris Est France

2 Outline o Performance Evaluation of Routing Protocols - Mobility/Speed Analysis - Scalability Analysis - Traffic Analysis o Modeling Routing Overhead of Reactive Protocols o Design Requirements for Routing Link Metrics o Min Hop-count o ETX-based Metrics - Expected Transmission Count (ETX) - Expected Link Performance (ELP) - Expected Throughput (ETP) o New Link Metric IBETX 1

3 Outline o Performance Evaluation of Routing Protocols - Mobility/Speed Analysis - Scalability Analysis - Traffic Analysis o Modeling Routing Overhead of Reactive Protocols o Design Requirements for Routing Link Metrics o Min Hop-count o ETX-based Metrics - Expected Transmission Count (ETX) - Expected Link Performance (ELP) - Expected Throughput (ETP) o New Link Metric IBETX 1

4 Routing Protocols for Wireless Networks o Reactive protocols - Reactive protocols perform RD and RM - Take an action when request arrives - More suitable for mobile scenarios - AODV, DSR, DYMO o Proactive protocols - Periodically exchange information - More scalable - More suitable for less dynamic networks - DSDV, FSR, OLSR 2

5 Reactive Protocols Mobility/Speed/Scalability/Traffic Analyses Protocol Performance Reasons AODV +++ DSR ++ DYMO + Provides multiple timely routes in RT Time association LLR quickly repairs link Multiple routes in RC Promiscuous listening mode. Highest speed and highest mobility, low convergence Not suitable for highly dynamic scenarios Absence of grat. RREPs Absence of supplementary maintenance strategies 3

6 Proactive Protocols Mobility/Speed Analysis Protocol Performance Reasons DSDV +++ OLSR ++ FSR + Delay advertisement for route stabilization (Periodic, Nghbr) Keep data for route settling time Periodic and triggers updates More suitable for static or less dynamic networks MPR-redundancy provides convergence in high mobility Periodic updates Suitable for mobile scenarios as compared to static scenarios Graded-frequency mechanism 4

7 Proactive Protocols Scalability Analysis Protocol Performance Reasons OLSR +++ FSR ++ More density, more optimization through MPRs Absence of trigger updates Retransmission reduction through MPRs Exchange link state messages by GF (<MPR) Absence of trigger updates Broadcasting messages only to neighbors; no flooding DSDV + Transmission of trigger updates utilize more bandwidth 5

8 Proactive Protocols Traffic Analysis Protocol Performance Reasons FSR +++ Exchange link state messages by GF Absence of trigger updates no flooding OLSR ++ Absence of trigger updates Retransmission reduction by MPRs (better in dense) DSDV + Transmission of trigger updates utilize more bandwidth 6

9 Outline o Performance Evaluation of Routing Protocols - Mobility/Speed Analysis - Scalability Analysis - Traffic Analysis o Modeling Routing Overhead of Reactive Protocols o Design Requirements for Routing Link Metrics o Min Hop-count o ETX-based Metrics - Expected Transmission Count (ETX) - Expected Link Performance (ELP) - Expected Throughput (ETP) o New Link Metric IBETX

10 Modeling Routing Overhead-Motivation o [Park06] define cost for routing overhead for reactive protocols; o No evaluation for C E and C T, o We define Energy Consumed per packet, o We define Time Consumed per packet, 7

11 Modeling Routing Overhead-Motivation o [sal10] considered: - only RREQ; - modeled only RD - evaluated energy-cost o We consider: - both RREQ, RREP and RERR overheads - both RD and RM - evaluate both energy and time-cost of RD and RM 8

12 Flooding o Flooding exchanges topological information in a network oplain flooding - Each received packet is (re)transmitted osuper flooding - A packet is (re)transmitted, if it has a shorter path omprs flooding - Only MPR nodes flood oblind flooding - Each packet is (re)transmitted for first time 9

13 ERS 10

14 ERS 10

15 ERS 10

16 ERS 10

17 ERS 10

18 ERS 10

19 ERS 10

20 ERS 10

21 ERS 10

22 ERS 10

23 ERS 10

24 ERS 10

25 ERS 10

26 ERS 10

27 ERS 10

28 Energy-cost for RD o Energy-cost for a single ring, o We consider routing overhead for ERS, o Energy-cost for RD, 11

29 Time-cost for RD o AODV and DYMO waiting_time = 2 NODE_TRAVERSAL_TIME (TTL_VALUE + TIME_OUT_BUFFER) o For DSR - τ = 30ms - For TTL=TTL x 2, τ= τ x 2 12

30 Route maintenance 13

31 Route maintenance 13

32 Route maintenance o S establishes path to D 13

33 Route maintenance o S establishes path to D o AODV and DYMO use HELLO messages to check connectivity o DSR uses link level feed-back 13

34 Route maintenance o S establishes path to D o AODV and DYMO use HELLO messages to check connectivity o DSR uses link level feed-back o Node 7 detects link break 13

35 Route maintenance o S establishes path to D o AODV and DYMO use HELLO messages to check connectivity o DSR uses link level feed-back o Node 7 detects link break o AODV starts LLR o LLR repairs the broken links o Reduces chance of route rediscovery o More suitable in large networks 13

36 Local Link Repair o S establishes path to D o AODV and DYMO use HELLO messages to check connectivity o DSR uses link level feed-back o Node 7 detects link break o AODV starts LLR o LLR repairs the broken links o Reduces chance of route rediscovery o More suitable in large networks 13

37 Local Link Repair o S establishes path to D o AODV and DYMO use HELLO messages to check connectivity o DSR uses link level feed-back o Node 7 detects link break o AODV starts LLR o LLR repairs the broken links o Reduces chance of route rediscovery o More suitable in large networks 13

38 Local Link Repair o S establishes path to D o AODV and DYMO use HELLO messages to check connectivity o DSR uses link level feed-back o Node 7 detects link break o AODV starts LLR o LLR repairs the broken links o Reduces chance of route rediscovery o More suitable in large networks 13

39 RERR in reactive protocols AODV o If LLR fails then send RERR messages to neighbors o A RERR message is broadcasted (if there are many precursors) o Or unicasted (if there is only 1 precursor) o a node send upto RERR_RATELIMIT (=10) RERR messages per second 14

40 RERR in reactive protocols DSR o Success of PS o Node returns a RERR to sender of packet o Then salvages packet o Node 7 is the originating node for RERR 14

41 RERR in reactive protocols DSR o Success of PS o Node returns a RERR to sender of packet o Then salvages packet o Node 7 is the originating node for RERR 14

42 RERR in reactive protocols DYMO o Broadcast RERR messages at link failure o MsgHdr.HopLimit is to NET_DIAMETER (=10) o MsgHdr.HopCnt is set to one 14

43 Re-RD in reactive protocols AODV (LLR) o RREQ_RETRIES = 2 to start new RD DSR (PS) o Source piggy-backs RERR with new RREQ o Initiating RREQ based on MaxMainRexmt=2 DYMO o Decision based on RREQ_TRIES =3 in DYMO to start new RD 14

44 Energy-cost of RM AODV DSR DYMO 15

45 Time cost for RM AODV DSR DYMO 16

46 Outline o Performance Evaluation of Routing Protocols - Mobility/Speed Analysis - Scalability Analysis - Traffic Analysis o Modeling Routing Overhead of Reactive Protocols o Design Requirements for Routing Link Metrics o Min Hop-count o ETX-based Metrics - Expected Transmission Count (ETX) - Expected Link Performance (ELP) - Expected Throughput (ETP) o New Link Metric IBETX

47 Routing Link Metric o To have appreciable performance from a wireless network, a routing protocol (RP) plays an important role o Key component of a RP is link metric (LM) o LM helps a RP first to find all possible end-toend paths and then selects the fastest path 17

48 Routing Link Metric o Minimum Hop-count (Min-HC) is the most popular and IETF standard metric and is appropriately used by wireless RP s o Min-HC rapidly finds new paths; where quality paths could not be found in due time due to higher rates of node mobility o Min-HC is the simplest to calculate, as it avoids any computational burden on RP 18

49 Designing a routing metric o Minimum-hop path o Balancing traffic load o Minimizing delay o Maximizing data delivery/aggregating bandwidth o Minimizing energy consumptions o Minimizing channel/interface switching 19

50 Designing a routing metric o Minimizing interference o Maximizing route stability o Maximizing fault tolerance/minimizing route sensitivity o Avoiding short and long lived loops oconsidering performance trade-offs ominimizing the Computational overhead 20

51 Computational Overhead 21

52 Computational Overhead 21

53 Computational Overhead 21

54 Throughput MD 22

55 Throughput ETX MD 22

56 Throughput ML ETX MD 22

57 Throughput InvETX ML ETX MD 22

58 Plane Flooding 23

59 MPR Flooding 23

60 OLSR-ETX/InvETX/ML 23

61 OLSR-ETX/InvETX/ML 23

62 OLSR-MD 23

63 End-to-end Delay MD 24

64 End-to-end Delay ETX MD 24

65 End-to-end Delay InvETX ETX MD 24

66 End-to-end Delay ML InvETX ETX MD 24

67 Outline o Performance Evaluation of Routing Protocols - Mobility/Speed Analysis - Scalability Analysis - Traffic Analysis o Modeling Routing Overhead of Reactive Protocols o Design Requirements for Routing Link Metrics o Min Hop-count o ETX-based Metrics - Expected Transmission Count (ETX) - Expected Link Performance (ELP) - Expected Throughput (ETP) o New Link Metric IBETX

68 ETX-based Metrics Issues Solution Links have different characteristics Link asymmetry Low bit-rate links effect faster links Bandwidth Congestion and Collisions Interference 25

69 ETX-based Metrics Issue(s) not considered Interference Bandwidth Link asymmetry Bandwidth and Interference Link asymmetry + Bandwidth + Interference Link metric ETT, DBETX, EETT, EstdTT, MCR, METX, ETP ELP, WCETT, MIC, iaware, EDR, ETX Dist MTM ETX, metx and ENT No Metric yet 26

70 Min Hop-count 27

71 Min Hop-count 27

72 Min Hop-count 27

73 Min Hop-count 27

74 Min Hop-count 27

75 Min Hop-count 27

76 Min Hop-count 27

77 ETX 28

78 ETX 28

79 ETX 28

80 ETX 28

81 ETX 28

82 ETX 28

83 ETX 28

84 ETX 28

85 ETX 28

86 ETX 28

87 ETX-based Metrics - ELP 29

88 ETX-based Metrics - ELP 29

89 ETX-based Metrics - ELP 29

90 ETX-based Metrics - ELP 29

91 ETX-based Metrics - ELP 29

92 ETX-based Metrics - ELP 29

93 ETX-based Metrics - ELP 29

94 ETX-based Metrics - ELP 29

95 ETX-based Metrics - ETP 30

96 ETX-based Metrics - ETP 30

97 ETX-based Metrics - ETP 30

98 ETX-based Metrics - ETP 30

99 ETX-based Metrics - ETP 30

100 ETX-based Metrics - ETP 30

101 Outline o Performance Evaluation of Routing Protocols - Mobility/Speed Analysis - Scalability Analysis - Traffic Analysis o Modeling Routing Overhead of Reactive Protocols o Design Requirements for Routing Link Metrics o Min Hop-count o ETX-based Metrics - Expected Transmission Count (ETX) - Expected Link Performance (ELP) - Expected Throughput (ETP) o New Link Metric IBETX

102 IBETX - Motivation o In a WMN-backbone, either stationary or minimally mobile nodes are interconnected o Thus, WMNs demand low end-to-end delay and high throughput paths from its operating RP o So, RP must choose a Quality LM to select the quality links 31

103 IBETX - Motivation 32

104 IBETX - Motivation 33

105 IBETX - Motivation 34

106 IBETX - Motivation 35

107 IBETX 36

108 IBETX 36

109 IBETX 36

110 IBETX 36

111 IBETX 36

112 IBETX 36

113 IBETX 36

114 IBETX 36

115 Simulation parameters Parameters Values NS version NS 2.34 Protocol DSDV Link metrics ETX, ELP, ETP, IBETX Network area 1000m x 1000m Number of nodes 50 Source-destination pairs 20 CBR packet size 640Bytes Data traffic rates 2pack/s to 10pack/s Topologies Mean of 5 random topologies is used Total simulation time 900s 37

116 Simulations-Throughput ETX 38

117 Simulations-Throughput ELP ETX 38

118 Simulations-Throughput ETP ELP ETX 38

119 Simulations-Throughput IBETX ETP ELP ETX 38

120 Simulations-Throughput IBETX 10% ETP ELP ETX 38

121 Simulations-Throughput IBETX 10% ETP 10% ELP ETX 38

122 Simulations-Throughput IBETX 10% ETP 10% ELP ETX 19% 38

123 Simulations-End-to-end Delay ETX 39

124 Simulations-End-to-end Delay ETX ELP 39

125 Simulations-End-to-end Delay ETX ELP ETP 39

126 Simulations-End-to-end Delay ETP IBETX ETX ELP 39

127 Simulations-End-to-end Delay ETX ELP ETP 15% IBETX 39

128 Simulations-End-to-end Delay ELP 16% ETX ETP 15% IBETX 39

129 Simulations-End-to-end Delay ELP 16% ETX 24% ETP 15% IBETX 39

130 Conclusion o Performance Evaluation and Comparison of: - 3 reactive protocols - 3 proactive protocols o Modeling cost paid by reactive routing protocols for RD and RM in the form of time-spent and energy-consumed o Performance study of ETX-based metrics o Designing a Quality Link Metric and InvETX o IBETX: Proposal and validation 40

131