Download presentation
Presentation is loading. Please wait.
Published byJessica Carr Modified over 9 years ago
1
1 Exploiting Antenna Capabilities in Wireless Networks Nitin Vaidya Electrical and Computer Engineering, and Coordinated Science Lab (CSL) University of Illinois at Urbana-Champaign www.crhc.uiuc.edu/wireless/
2
2 Wireless Capacity Wireless capacity limited In dense environments, performance suffers How to improve performance?
3
3 Improving Per-Flow Capacity
4
4 Add Spectrum Multi-channel versions of IEEE 802.11 Practical limits on how much spectrum may be used
5
5 Power Control to Improve Spatial Reuse ABCD ABCD
6
6 Improving Communication Locality Local communication (among nearby nodes) uses less “space” Allows spatial reuse among different flows Improves per-flow capacity Not always feasible: Application-dependent
7
7 Exploit Infrastructure Infrastructure provides a “tunnel” through which packets can be forwarded Can effectively improve locality of communication Infrastructure access can become a bottleneck E A BS1BS2 X Z infrastructure Ad hoc connectivity
8
8 Improving Per-Flow Capacity Previous techniques are all useful, but have limitations Dense networks likely to require further improvements in capacity Exploit other forms of diversity Mobility Antennas
9
9 Exploiting Antennas
10
10 Antennas: Many Possibilities Directional antennas Diversity antennas Reconfigurable antennas …
11
11 Exploiting Antennas Need protocol adaptations to exploit available antenna capabilities Not sufficient to modify physical layer alone Higher layer adaptation often necessary: medium access control (MAC) and routing
12
12 This Talk Protocols for Ad Hoc Networks using Directional Antennas Issues of interest Medium access control Neighbor discovery Routing Longer links, shorter routes Longer times to failure Broadcast-based discovery harder This talk Deafness problem MAC-Layer Anycasting
13
13 Outline Preliminaries A simple MAC protocol and the “deafness” problem MAC-layer anycasting
14
14 Ad Hoc Networks Formed by wireless hosts which may be mobile Without necessarily using a pre-existing infrastructure Routes between nodes may potentially contain multiple hops Hidden terminals
15
15 Antenna Model 2 Operation Modes: Omni & Directional Directional mode typically has sidelobes Not all antennas represented by this model
16
16 Antenna Model Omni Mode: Omni Gain = Go Directional Mode: Capable of beamforming in specified direction Directional Gain = Gd (Gd > Go) Received power Transmit power * G tx * G rx
17
17 Benefits of Directional Antennas Greater Received Power Longer links may be formed B A C D May lower Tx power, reducing interference to others
18
18 Benefits of Directional Antennas Low gain in unwanted directions Reduces interference to others Example ….
19
19 Using Omni-directional Antennas When C receives from D, B cannot transmit C B A D
20
20 Using Directional Antennas C may receive from D, and simultaneously B may transmit to A C B A D
21
21 A detour …
22
22 ABC Hidden Terminal Problem Node B can communicate with A and C both A and C cannot hear each other When A transmits to B, C cannot detect the transmission using the carrier sense mechanism If C transmits, collision may occur at node B
23
23 RTS/CTS Handshake in 802.11 Sender sends Ready-to-Send (RTS) Receiver responds with Clear-to-Send (CTS) RTS and CTS announce the duration of the transfer Nodes overhearing RTS/CTS keep quiet for that duration D C BA CTS (10) RTS (10) 10
24
24 Outline Preliminaries A simple MAC protocol and the “deafness” problem MAC-layer anycasting
25
25 Directional MAC (DMAC) Idle node listens in omni-directional mode Sender sends a directional RTS towards intended receiver Receiver responds with directional CTS
26
26 Directional MAC (802.11 Variant) DATA and ACK transmitted and received directionally Nodes overhearing RTS or CTS remember not to transmit in corresponding directions Overhearing nodes may transmit in other directions
27
27 Directional MAC C remembers not to transmit in A’s direction C may transmit towards D D A C B RTS
28
28 Issues with DMAC Hidden terminals due to asymmetry in gain A does not get RTS/CTS from C/B C A B Data RTS A’s RTS may interfere with C’s reception of DATA
29
29 Issues with DMAC: Deafness Deafness: C does not know why no response from A Cannot differentiate between collision, and busy node A Conservative response is to “backoff” and try later D AB C RTS
30
30 Illustration B initiates communication to A While A is busy, C transmits RTS to A No response from A C waits a while, tries again No response, C waits longer … When A becomes free, C in wait mode A become busy again, …. Repeat AB C RTS
31
31 RTS Backoff Data RTS CTS ACK Data CTS RTS B initiates communication to A While A is busy, C transmits RTS to A No response from A C waits a while, tries again No response, C waits longer … When A becomes free, C in wait mode A become busy again, …. Repeat Illustration Packet drop AB C
32
32 Impact of Deafness Unnecessary transmissions of RTS Increased packet drops Increased delay and variance Unfairness among flows
33
33 Solutions to Deafness Deafness since C does not know A is busy Make C aware that A is busy Require A to transmit a busy signal while receiving Alternative: A transmits a “free” signal after it become idle RTS Backoff Data RTS CTS ACK Data CTS RTS Packet drop AB C
34
34 Solution: Tone DMAC Nodes unable to communicate with A adapt backoff based on the “tone” from A Think of it as “free-tone” as opposed to a “busy- tone” A node need only use tone or data channel at any time, not both RTS Backoff Data RTS CTS ACK AB C Tone RTS CTS Data Backoff
35
35 Tone DMAC Why a narrow-band tone? Save bandwidth Trade-off Narrow-band signal prone to fading: Use long enough tone duration Aliasing, since C cannot tell who transmitted a tone –Use multiple tones –One tone per node too expensive –Share tones
36
36 Tone DMAC Node i transmit tone f i for duration t i f i and t i functions of the node identifier i f i = i mod F t i = i mod T
37
37 Tone DMAC When a node, such as C in our example, hears a tone f for duration t, node C determines whether the tone could have been sent by its intended traget (node A in our example) If C determines that A is the tone sender, C reduces its waiting time before next RTS Aliasing can occur since multiple nodes can hash to the same tuple { f, t }
38
38 Tone DMAC Example
39
39 Backoff: Two flows to common receiver Another possible improvement: Backoff Counter for DMAC flows Backoff Counter for ToneDMAC flows time Backoff Values
40
40 Packet Drops: Three flows, common receiver DMAC ToneDMAC time
41
41 UDP Throughput: Multiple multihop flows ToneDMAC outperforms DMAC, ZeroToneDMAC ZeroToneDMAC = DMAC with only omnidirectional Backoff
42
42 Delay Performance: 2 flows, common Rx Large fluctuation in DMAC packet delay Higher variance
43
43 TCP Throughput: Multiple multihop flows RTT estimation of TCP better with ToneDMAC due to low delay variance
44
44 DMAC Summary Deafness aggrevated by directional communication “Free” tones, or other alternative mechanisms, appear useful to reduce degradation caused by deafness Practicality issue: Tone assignment Fading Topic of ongoing research
45
45 MAC-Layer Anycasting
46
46 Observation Network layer typically selects one “optimal” route MAC layer required to forward packet to next hop neighbor on this route “Optimal” route selection based on a long-term view of the network Independent of instantaneous channel conditions at each hop
47
47 Improvement ? MAC layer aware of local link conditions Congestion, channel fluctuations at smaller time scale Power constraints for transmission Virtual carrier sensing information (NAV in 802.11) Exploit MAC layer awareness Especially when using directional antennas Forward packets based on combination of Long-term directives of routing layer, and Short-term knowledge at MAC layer
48
48 Our Proposal Make forwarding decisions at the MAC layer Utilize information already available to the MAC layer (as opposed to explicitly gathering feedback) With DMAC, a node already knows that it cannot transmit in certain directions Our approach can be combined with mechanisms that gather information explicitly
49
49 MAC-Layer Anycasting Source often has multiple “good” routes to sink Typically, one random downstream neighbor chosen Supply multiple downstream neighbors to MAC layer MAC layer chooses any one of the neighbors based on available information, and unicasts the packet
50
50 MAC-Layer Anycast Framework Anycast module receives group of downstream neighbors Anycast group = {A, B, X} Anycast module forms anycast sequence (based on chosen policy) Seq. = {X, X, B, A, X, B, A} MAC layer attempts to transmit to “available” neighbors Network Layer MAC Layer Physical Layer Anycast Module
51
51 Directional MAC X DS DRTS Y
52
52 Directional MAC X DS DCTS Remember to not transmit towards D Y
53
53 MAC Constraints Route from S to D: {S,A,B,D} Assume A communicating with B S cannot send packet to A Multiple retransmissions can be avoided by forwarding packet to X instead Specify anycast group specified as {A, X} A S Y D B X Directional Beam Patterns
54
54 DNAV Constraints Communication between E and F requires S to set DNAV in direction of E Communication between S and A not possible until E completes transmission Communication between S and X may be possible Anycasting with group {A,X} can improve performance F E X A S
55
55 Not Allowed DNAV Constraints F E X S A Communication between E and F requires S to set DNAV in direction of E Communication between S and A not possible until E completes transmission Communication between S and X may be possible Anycasting with group {A,X} can improve performance
56
56 DNAV Constraints F E XS A Allowed Communication between E and F requires S to set DNAV in direction of E Communication between S and A not possible until E completes transmission Communication between S and X may be possible Anycasting with group {A,X} can improve performance
57
57 MAC Constraints – Omni Antennas Route from S to D: {S,A,B,D} While F communicating to E, A is silenced by CTS from E S transmits RTS to A, receives no reply, retransmits Multiple retransmission can be avoided by forwarding packet to X Anycast group specified to S can be {A, X}
58
58 Power Constraints RT P N With PCMA, node R announces additional interference that it can tolerate To initiate communication to N, T must choose power level according to this tolerance Power level to transmit to N is too high. However, transmission to P is feasible MAC-Layer anycasting can forward packets with PCMA. Anycast group {P, N}
59
59 Power Constraints RT P N With PCMA, node R announces additional interference that it can tolerate To initiate communication to N, T must choose power level according to this tolerance Power level to transmit to N is too high. However, transmission to P is feasible MAC-Layer anycasting can forward packets with PCMA. Anycast group {P, N}
60
60 Power Constraints RT P N With PCMA, node R announces additional interference that it can tolerate To initiate communication to N, T must choose power level according to this tolerance Power level to transmit to N is too high. However, transmission to P is feasible MAC-Layer anycasting can forward packets with PCMA. Anycast group {P, N}
61
61 Design Issues and Tradeoffs
62
62 “Digression” Anycasting can bypass unavailable links Each intermediate node locally performs anycasting Local (greedy) decisions can cause Route to digress significantly from global optimal Need to restrict digression below tolerance
63
63 Digression Say, Anycast group = Neighbors on the minimum and (minimum+1)-hop routes {S,X,J,P,K,Z,D} digresses 3 hops more that {S,A,B,D}
64
64 Out-of-Order Delivery MAC-Layer anycasting performed on per-packet basis Delay on the different routes can be different Out of order packet delivery possible TCP-like transport protocols may encounter problems
65
65 Source Routing Source routing – source specifies all possible routes To perform anycasting with source routing Source includes enough information for intermediate nodes to form anycast group Possible implementation – include a directed acyclic graph (DAG) Including DAG in packet – larger control overhead
66
66 Preliminary Evaluation (Anycasting)
67
67 Grid topology, 5 flows, 3 hops
68
68 Large Grid topology, 10 flows, 5 hops
69
69 Anycast: Summary MAC-Layer anycasting can improve performance Several tradeoffs arise On-going work
70
70 Conclusion Directional antennas can benefit performance But need suitable protocols On-going work: Cheaper antennas that can improve performance Testbed deployment
71
71 Thanks! www.crhc.uiuc.edu/wireless Acknowledgements Romit Roy Choudhury, UIUC Ram Ramanathan, BBN Xue Yang, UIUC
72
72 Another Problem Performing directional carrier sensing when in wait mode leads to another instance of deafness While C waits to transmit to A, it beamforms and performs carrier sensing C cannot hear RTS from D AB C RTS D
73
73 Solutions to Deafness Nodes required to switch to omni mode during back-off C can hear D while waiting for A Trade-off: C may receive transmission from E to F, and not be able to receive from D, or transmit to A AB C RTS D E
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.