Transcript TCP for Mobile and Wireless Hosts

Directional Antennas
in
Ad Hoc Networks
Nitin Vaidya
University of Illinois at Urbana-Champaign
Joint work with
Romit Roy Choudhury, UIUC
Xue Yang, UIUC
Ram Ramanathan, BBN
1
Mobile 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
2
Mobile Ad Hoc Networks

May need to traverse multiple links to reach a
destination
3
Mobile Ad Hoc Networks (MANET)

Mobility causes route changes
4
Why Ad Hoc Networks ?

Potential ease of deployment

Decreased dependence on infrastructure
5
Many Applications




Personal area networking
cell phone, laptop, ear phone, wrist watch
Military environments
soldiers, tanks, planes
Civilian environments
taxi cab network
meeting rooms
sports stadiums
boats, small aircraft
Emergency operations
search-and-rescue
policing and fire fighting
6
Many Variations

Fully Symmetric Environment
all nodes have identical capabilities and responsibilities

Asymmetric Capabilities
transmission ranges and radios may differ
battery life at different nodes may differ
processing capacity may be different at different nodes

Asymmetric Responsibilities
only some nodes may route packets
some nodes may act as leaders of nearby nodes (e.g.,
cluster head)
7
Many Variations

Traffic characteristics may differ in different ad hoc
networks
bit rate
timeliness constraints
reliability requirements
unicast / multicast / geocast
host-based addressing / content-based addressing /
capability-based addressing

May co-exist (and co-operate) with an infrastructurebased network
8
Many Variations

Mobility patterns may be different

Mobility characteristics
people sitting at an airport lounge
New York taxi cabs
kids playing
military movements
personal area network
speed
predictability
• direction of movement
• pattern of movement
uniformity (or lack thereof) of mobility characteristics among
different nodes
9
Challenges


Limited wireless transmission range
Broadcast nature of the wireless medium
– Hidden terminal problem






Packet losses due to transmission errors
Mobility-induced route changes
Mobility-induced packet losses
Battery constraints
Potentially frequent network partitions
Ease of snooping on wireless transmissions (security
hazard)
10
Question

Can ad hoc networks benefit from the progress made
at physical layer ?
Efficient coding schemes
Power control
Adaptive modulation
Directional antennas
…

Need improvements to upper layer protocols
11
Directional Antennas
12
Using Omni-directional Antennas
A Frozen
node
D
B
S
A
13
Directional Antennas
Not possible
using Omni
D
B
S
C
A
14
Comparison
Omni
Directional
Spatial Reuse
Low
High
(varies inversely with
beamwidth)
Connectivity
Low
High
Interference
Omni
Directional
Cost & Complexity
Low
High
15
Questions

Are Directional antennas beneficial in ad hoc
networks ?
To what extent ?
Under what conditions ?
16
Research Direction

Identify issues affecting directional communication

Evaluate trade-offs across multiple layers

Design protocols that effectively use directional
capabilities
Caveat: Work-in-Progress
17
Preliminaries
18
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
A
B
C
19
RTS/CTS Handshake




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
C
10
RTS (10)
A
B
CTS (10)
D
10
20
IEEE 802.11




Physical carrier sense
Virtual carrier sense using Network Allocation Vector
(NAV)
NAV is updated based on overheard
RTS/CTS/DATA/ACK packets, each of which
specified duration of a pending transmission
Nodes stay silent when carrier sensed busy
(physical/virtual)
21
Antenna Model
22
Antenna Model

2 Operation Modes: Omni & Directional
23
Antenna Model

Omni Mode:
Omni Gain = Go
Idle node stays in Omni mode.

Directional Mode:
Capable of beamforming in specified direction
Directional Gain = Gd (Gd > Go)
24
Directional Neighborhood
A
B
C
A and B are Directional-Omni (DO) neighbors
B and C are Directional-Directional (DD) neighbors
25
A Simple Directional MAC Protocol
(DMAC)
26
DMAC Protocol

A node listens omni-directionally when idle
Only DO links can be used

Sender node sends a directional-RTS using specified
transceiver profile

Receiver of RTS sends directional-CTS
27
DMAC Protocol

DATA and ACK transmitted and received directionally

Nodes overhearing RTS or CTS sets up NAV for that
DOA (direction of arrival)

Nodes defer transmitting only in directions for which
NAV is set
28
Directional NAV (DNAV)

Node E remembers directions in which it has
received RTS/CTS, and blocks these directions.

Transmission initiated only if direction of transmission
does not overlap with blocked directions.
29
Directional NAV (DNAV)

E has DNAV set due to RTS from H. Can talk to B
since E’s transmission beam does not overlap.
30
Example
C
E
D
B
B and C communicate
D & E cannot: D blocked with DNAV
A
D and A communicate
31
Issues with DMAC

Hidden terminals due to asymmetry in gain
A does not get RTS/CTS from C/B
Data
RTS
A
B
C
A’s RTS may interfere with C’s reception of DATA
32
Problems with DMAC

Hidden terminals due to directionality
Due to unheard RTS/CTS
D
B
A
C
A beamformed in direction of D
 A does not hear RTS/CTS from B/C
A may now interfere at C
33
Issues with DMAC: Deafness
• Deafness
Z
RTS
A
B
DATA
RTS
Y
RTS
X
X does not know node A is busy.
X keeps transmitting RTSs to node A
With 802.11 (omni antennas), X would be aware that A is busy, and
defer its own transmission
34
Problems with DMAC

Shape of Silenced Regions
Region of interference for
omnidirectional transmission
Region of interference for
directional transmission
35
Problems with DMAC

Since nodes are in omni mode when idle, RTS
received with omni gain

DMAC can use DO links, but not DD links
A
B
C
36
DMAC Trade-off

Benefits
Better Network Connectivity
Spatial Reuse

Disadvantages
– Increased hidden
terminals
– Deafness
– Directional interference
– Uses only DO links
37
Solving DMAC Problems

Are improvements possible to make directional MAC
protocols more effective ?

One possible improvement: Use DD links
38
Using DD Links

Possible to exploit larger range of directional
antennas.
A
C
A & C are DD neighbors, but cannot communicate with DMAC
If A & C could be made to point towards each other, single hop
communication may be possible
39
Multi-Hop RTS: Basic Idea
D
C
A
B
DO neighbors
E
DD neighbors
F
G
A source-routes RTS to D through adjacent DO neighbors
(i.e., A-B-C-D)
When D receives RTS, it beamforms towards A, forming a
DD link.
40
MMAC protocol
D
E
F
H
C
A
B

A transmits RTS in the direction of its DD neighbor,
node D
Blocks H from communicating in the direction H-D

A then transmits multi-hop RTS using source route

A beamforms towards D and now waits for CTS
G
41
D
MMAC protocol
E
F
H
C
A
B
G

D receives MRTS from C and transmits CTS in the direction of A
(its DD neighbor).

A initiates DATA communication with D

H, on hearing RTS from A, sets up DNAVs towards both H-A
and H-D. Nodes B and C do not set DNAVs.

D replies with ACK when data transmission finishes.
42
Performance

Simulation
Qualnet simulator 2.6.1
CBR traffic
Packet Size – 512 Bytes
802.11 transmission range = 250 meters.
Channel bandwidth 2 Mbps
Mobility - none
43
Impact of Topology
• Nodes arranged in
linear configurations
reduce spatial reuse for
directional antennas
44
Impact of Topology
IEEE 802.11 = 1.19 Mbps
DMAC = 2.7 Mbps
IEEE 802.11 = 1.19 Mbps
DMAC = 1.42 Mbps
45
“Aligned” Flows
MMAC
802.11
DMAC
46
“Unaligned” Flows
MMAC
802.11
DMAC
47
“Unaligned” Flows & Topology
MMAC
802.11
DMAC
48
Delay: “Unaligned” Flows & Topology
49
Directional MAC: Summary

Directional MAC protocols can improve throughput
and decrease delay
But not always

Performance dependent on topology
50
Routing using Directional Antennas
51
Motivation

Directional antennas affect network layer, in addition
to MAC protocols
52
Dynamic Source Routing [Johnson]

Sender floods RREQ through the network

Nodes forward RREQs after appending their names

Destination node receives RREQ and unicasts a
RREP back to sender node, using the route in which
RREQ traveled
53
Route Discovery in DSR
Y
Z
S
E
F
B
C
M
J
A
L
G
H
K
I
D
N
Represents a node that has received RREQ for D from S
54
Route Discovery in DSR
Y
Broadcast transmission
[S]
S
Z
E
F
B
C
M
J
A
L
G
H
K
I
D
N
Represents transmission of RREQ
[X,Y]
Represents list of identifiers appended to RREQ
55
Route Discovery in DSR
Y
Z
S
E
[S,E]
F
B
C
A
M
J
[S,C]
H
G
K
I
L
D
N
56
Route Discovery in DSR
Y
Z
S
E
F
B
[S,E,F]
C
M
J
A
L
G
H
I
[S,C,G] K
D
N
• Node C receives RREQ from G and H, but does not forward
it again, because node C has already forwarded RREQ once
57
Route Discovery in DSR
Y
Z
S
E
[S,E,F,J]
F
B
C
M
J
A
L
G
H
D
K
I
[S,C,G,K]
N
• Nodes J and K both broadcast RREQ to node D
58
Route Reply in DSR
Y
Z
S
E
RREP [S,E,F,J,D]
F
B
C
M
J
A
L
G
H
K
I
Represents RREP control message
D
N
59
DSR over Directional Antennas

RREQ broadcast by sweeping
To use DD links
60
Route Discovery in DSR
Y
Z
S
E
[S,E,F,J]
F
B
C
M
J
A
L
G
H
D
K
I
[S,C,G,K]
N
• Nodes J and K both broadcast RREQ to node D
61
Trade-off
Larger Tx Range
Few Hop Routes
Fewer Hop Routes
Low Data Latency
Smaller Angle
More Sweeping
High Sweep Delay
High Overhead
62
Route discovery latency … Single flow, grid
topology (200 m distance)
DDSR4
DDSR6
DSR
63
Observations

Advantage of higher transmit range significant only at
higher distance of separation.

Grid distance = 200 m --- thus no gain with higher tx
range of DDSR4 (350 m) over 802.11 (250 m).
However, DDSR4 has sweeping delay. Thus route discovery
delay higher
64
Throughput
DDSR18
DDSR9
DSR
Sub-optimal routes chosen by DSR because destination
node misses the shortest RREQ, while beamformed.
65
Route Discovery in DSR
F
J
RREP
J
D
K
RREQ
N
D receives RREQ from J, and replies with RREP
D misses RREQ from K
66
Delayed RREP Optimization

Due to sweeping – earliest RREQ need not have
traversed shortest hop path.
RREQ packets sent to different neighbors at different points
of time

If destination replies to first arriving RREP, it might
miss shorter-path RREQ

Optimize by having DSR destination wait before
replying with RREP
67
Routing Overhead

Using omni broadcast, nodes receive multiple copies
of same packet - Redundant !!!
• Broadcast Storm Problem

Using directional Antennas – can do better ?
68
Routing Overhead
Use K antenna elements to forward broadcast packet.
K = N/2 in simulations
Footprint
of Tx
Ctrl Overhead 
 (No. Ctrl Tx)  (Footprint of Tx)
=
 No. Data Packets
69
Routing Overhead
Control overhead reduces
Beamwidth of antenna element (degrees)
70
Directional Antennas over mobile scenarios

Frequent Link failures
Communicating nodes move out of transmission range

Possibility of handoff
Communicating nodes move from one antenna to another
while communicating
71
Directional Antennas over mobile scenarios

Link lifetime increases using directional antennas.
Higher transmission range - link failures are less frequent

Handoff handled at MAC layer
If no response to RTS, MAC layer uses N adjacent antenna
elements to transmit same packet
Route error avoided if communication re-established.
72
Aggregate throughput over random mobile
scenarios
DDSR9
DSR
73
Observations

Randomness in topology aids DDSR.

Voids in network topology bridged by higher
transmission range (prevents partition)

Higher transmission range increases link lifetime –
reduces frequency of link failure under mobility

Antenna handoff due to nodes crossing antenna
elements – not too serious
74
Conclusion

Directional antennas can improve performance

But suitable protocol adaptations necessary

Also need to use suitable antenna models

… plenty of problems remain
75
Thanks!
www.crhc.uiuc.edu/~nhv
76
77
Adaptive Modulation
Joint work with Gavin Holland and Victor Bahl
78
Adaptive Modulation

Channel conditions are time-varying

Received signal-to-noise ratio changes with time
A
B
79
Adaptive Modulation


Multi-rate radios are capable of transmitting at
several rates, using different modulation schemes
Choose modulation scheme as a function of channel
conditions
Modulation schemes provide
a trade-off between
throughput and range
Throughput
Distance
80
Adaptive Modulation

If physical layer chooses the modulation scheme
transparent to MAC
MAC cannot know the time duration required for the transfer

Must involve MAC protocol in deciding the
modulation scheme
Some implementations use a sender-based scheme for this
purpose [Kamerman97]
Receiver-based schemes can perform better
81
Sender-Based “Autorate Fallback”
[Kamerman97]

Probing mechanisms

Sender decreases bit rate after X consecutive
transmission attempts fail

Sender increases bit rate after Y consecutive
transmission attempt succeed
82
Autorate Fallback

Advantage
Can be implemented at the sender, without making any
changes to the 802.11 standard specification

Disadvantage
Probing mechanism does not accurately detect channel
state
Channel state detected more accurately at the receiver
Performance can suffer
Since the sender will periodically try to send at a rate
higher than optimal
Also, when channel conditions improve, the rate is not
increased immediately
•
•
83
Receiver-Based Autorate MAC
[Holland01mobicom]

Sender sends RTS containing its best rate estimate

Receiver chooses best rate for the conditions and
sends it in the CTS

Sender transmits DATA packet at new rate

Information in data packet header implicitly updates
nodes that heard old rate
84
Receiver-Based Autorate MAC Protocol
C
A
RTS (2 Mbps)
B
CTS (1 Mbps)
Data (1 Mbps)
D
NAV updated
using rate
specified in the
data packet
85
Extra slides
86
Directional Antennas in Random Topologies
Higher transmission range improves connectivity in
addition to achieving fewer hop routes.
E.g. Link a-b not possible using Omni transmission.
87
Effect of Beamwidth in Random Static
Topologies
88