SlideShare a Scribd company logo
1 of 8
Download to read offline
1
1
Medium access control in Wireless Ad-hoc Networks
Presented by
Jin Xu and Lan Nguyen
2
Outline
Classifications of MAC protocols
Five Phase Reservation Protocol (FPRP)
Distributed Wireless Ordering Protocol (DWOP)
Conclusions
3
Classifications of MAC protocols
Contention-based
A node contends with its neighbors to access the channel
No QoS guarantees
MACAW, FAMA, BTMA
Contention-based with reservation
Reserve bandwidth a priori
Can provide QoS support to real time traffic
D-PRMA, CATA, FPRP
Contention-based with scheduling
Focus on transmission scheduling of the nodes
Fair and no starvation
DPS, DWOP, DLPS
4
Five-Phase Reservation Protocol (FPRP)
Contention-based with reservation
Single channel time division multiple access
Fully distributed (synchronized)
Slot reservation using a 5 phase process
Parallel
Localized process
Scalable (insensitive to network size)
5
FPRP: the model
Nodes keep perfect timing
A link between 2 nodes is noiseless, symmetric
The network topology not change when FPRP is
performed
When multiple packets arrive at a node, they are
destroyed
A node can tell whether 0, 1, or multiple packets are
transmitted when in receiving mode
Every node has a unique ID
6
FPRP: overview
Time is divided into frames
2
7
FPRP: 5 phase
Reservation request (RR)
Collision report (CR)
Reservation confirmation (RC)
Reservation acknowledgement (RA)
Packing and elimination (P/E)
A node keeps global timing, and knows when a 5-phase
cycle starts.
A node can transmit or receive, but cannot do both at
the same time.
8
Phase 1: Reservation request
A node which wants to make a reservation sends a
Reservation Request packet (RR) with probability p
Other nodes listen
Requesting
Node
9
Phase 2: Collision report
If a node receives multiple RR’s in phase 1, it
transmits a Collision Report packet (CR)
o.w. silent
Requesting node (RN) transmission node (TN)
10
Phase 3: Reservation confirmation
TN sends Reservation Confirmation packet (RC)
Every node (1 hop away) which receives RC know
the slot has been reserved cease contention,
receiving
11
Phase 4: Reservation acknowledgement
A node ack a RC just received
by sending a Reservation Ack
packet (RA)
Inform nodes 2 hops away
Not transmitting
Prevent isolated node from
transmitting
Resolve isolated deadlock
12
Phase 5: packing / elimination
Every node 2 hops away from TN sends a Packing
packet (PP)
A node receiving PP learns there is a TN 3 hops away
Adjust its contention prob.
to reuse time slot efficiently
TN sends an Elimination packet (EP) with a
probability of 0.5
Attempt to resolve a non-isolated deadlock
3
13
An example
14
How to calculate the contention probability
nc : # of nodes that contend within 2 hops
nb : # of nodes within 2 hops that have to contend
in next slot due to a nearby success (cannot contend
in current slot)
R1: a portion of 1 hop neighbors from success cease
to contend in the current slot
R2: 2 hop neighbors
R3: 3 hop neighbors
15
How to calculate the contention probability
At the beginning of a reservation slot, a node resets
its nc and nb:
nc = nb nb = 0
After every reservation cycle, on hearing an:
Idle: nc = nc – 1
Collision: nc = nc + 1 / (e - 2)
Success:
0 hop: done;
1 hop: nc = nc – 1 nb = nb + nc R1 nc =nc(1 – R1)
2 hops: nc = nc – 1 nb = nb + nc R2 nc =nc(1 – R2)
3 hops: nb = nb + nc R3 nc =nc(1 – R3)
P = 1/nc
16
Simulation results
The # of nodes is 100, 200, 300 and 400 from bottom to top
17
Effects of nodal mobility
The observation time t is 0, 0.5, 1, 2, 4, 8, 16, 32, 64, 128
seconds from bottom to top
BM RCSt = 0
18
Protocol considerations and applications
Time synchronization issue:
GPS can provide accurate global time
Some applications:
TDMA schedule produced can be used to transmit user
generated packets
Make reservations for network control traffic
Can provide good service for multimedia (voice) traffic
4
19
Conclusions for FPRP
Fully distributed, requiring no a prior knowledge
about the network
Generate transmission schedules with low amount of
overhead
Not affected much by the network size and nodal
mobility
Suitable for use in large, mobile ad hoc network
20
Problems of IEEE 802.11
Unfair channel access
Due to random access nature of wireless 802.11
Hard to support QoS (priority routing/scheduling
e.g. diverges significantly from FIFO order
Even worse in more complex topologies, e.g.
Asymmetric information
Perceived collision
21
Asymmetric information
Symmetric information sharing
All nodes are within radio range of each other,
hence can hear all the RTS and CTS
All nodes have the same probability of accessing
the channel
Asymmetric information sharing (opposite)
All nodes are not within radio range of each other
Nodes have unequal channel access probabilities
22
Example
Node 1 does not know about
flow B, so it has to send RTS
randomly
Node 2 may not be able to
send CTS due to either:
Deferral for flow B transmission
Didn’t receive RTS due to
collision with flow B
Node 3 knows about flow A
via node 2, hence it knows
the “right time” to send RTS
Flow A: 5%
Flow B: 95%
23
A possible solution
MACAW
Node 2 sends Request-for-RTS
(RRTS) to node 1
Upon receiving RRTS, node 1
sends RTS immediately
Works only if node 2 has
received RTS
Not included in IEEE 802.11
24
Perceived collision
In the previous example, flow B has info
about flow A (not the vice versa) -> flow B
gets more channel access
But this is not always the case
As knowing more information about other
flows makes a flow defer access to more
flows
5
25
Perceived collision (cont.)
Flows A and C can access the channel simultaneously
Flow B contains information about both flows A & C,
i.e. it has to wait for both flows A & C
Flow A: 36%
Flow B: 28%
Flow C: 36%
26
DWOP protocol
Goals
FIFO-like behavior
Piggy-backing arrival times
Modifying IEEE 802.11
Receiver participation
Stale entry detection
27
Goals of DWOP
Providing fair channel access
Scheduling packets in the order that
approximates a reference scheduler
(e.g. FIFO)
28
FIFO-like behavior
Wireless contention for the channel access requires a
scheduler for all the contending nodes
For FIFO, packet priority is packet’s arrival time
Contending nodes share the arrival times
A scheduling table is needed
29
Piggy-backing the arrival times
A sender (receiver) piggy-
backs the arrival time of its
current (highest priority)
packet into RTS (CTS)
A sender (receiver) also
piggy-backs the arrival time
of its next highest priority
packet into the DATA (ACK)
packet’s transmission
30
Piggy-backing (cont.)
Nodes overhearing RTS,
CTS, DATA, ACK packets add
the attached arrival time into
their sorted scheduling
tables along with their own
packets’ arrival times
Also, nodes remove the
arrival time entry when
overhearing the completion
of DATA and ACK
6
31
Modifying IEEE 802.11
If the channel is busy,
behave like IEEE 802.11
If the channel is idle,
check the scheduling
table
If its packet has the
highest priority, send
RTS
Else defer as in IEEE
802.11
32
Does it work all the time ?
Only if all nodes are within radio range of each other
If not -> asymmetric information
Node 1 is not aware of flow B -> behaves like IEEE 802.11 -
> try to contend the access continuously and randomly
Flow B knows the arrival time of flow A -> defers if it has
lower priority -> less aggressive than flow A -> Flow B gets
less bandwidth
33
Receiver participation
Receiver of A (node 2)
know the arrival time of
flow B
“warn” node 1
include an out-of-order
notification in CTS/ACK
Upon receiving the out-
of-order notification
Node 1 finishes it current
transmission
Node 1 goes into a
backoff
34
Receiver participation (cont.)
Node 1 is allowed to complete its current out-of-order
transmission
Only approximate FIFO is achieved
DWOP can be modified to achieve perfect FIFO
But tradeoff between perfect FIFO scheduling and
network utilization
35
Stale entry elimination
Stale arrival time entry occurs in a node if:
Does not receive ACK (due to collision)
Stale entry detection
If there is a deletion below its packet’s position in the
scheduling table
Reaction
Remove the first entry in the scheduling table
36
Simulation experiments
Experiment setup
ns-2 simulator with cmu-wireless extension
Data packet size: 1000 bytes
Channel bandwidth: 2Mbps
CBR flows
Three topologies:
Asymmetric information topology
Perceived collision topology
More complex topology
7
37
Simulation experiments (cont.)
Performance metrics
Fair channel access
FIFO-like behavior
Ideal FIFO
Switch to another flow after one packet
transmission
38
Asymmetric information
39
Perceived collision topology
40
More complex topology
41
Summary of results
DWOP has less deviation from FIFO
DWOP: deviation bounded up to 4 packets
IEEE 802.11: unbounded deviation
DWOP achieves better fairness of
accessing the channel than IEEE 802.11
42
Limitations of DWOP
Flows of one single hop
Fixed packet size
FIFO scheduler
Total number of received packets is about 1/3
less than IEEE 802.11
Does not consider
Mobility
Channel errors (ECF [3])
Variable packet size
Other reference schedulers
8
43
DWOP conclusions
IEEE 802.11 does not take care of fair
channel access well
IEEE 802.11 diverges significantly from FIFO
DWOP provides fairness of channel access
DWOP approximates the FIFO order better
DWOP can be applied to other schedulers
QoS can be integrated in DWOP
44
Reference
[1]. A Five-Phase Reservation Protocol (FPRP) for Mobile Ad Hoc
Networks, C Zhu, M S Corson, Proceedings of IEEE INFOCOM,
1998.
[2]. Ordered Packet Scheduling in Wireless Ad Hoc Networks:
Mechanisms and Performance Analysis, V. Kanodia, C. Li, A.
Sabharwal, B. Sadeghi, Proceedings of ACM MOBIHOC, 2002.
[3]. Fair scheduling in wireless ad-hoc networks of location
dependent channel errors, Chen J., Somani A. K., Proceedings
of the 2003 IEEE International, 2003.
45
Question ?

More Related Content

What's hot

ALOHA Protocol (in detail)
ALOHA Protocol (in detail)ALOHA Protocol (in detail)
ALOHA Protocol (in detail)Hinal Lunagariya
 
Comprehensive survey on routing protocols for IoT
Comprehensive survey on routing protocols for IoTComprehensive survey on routing protocols for IoT
Comprehensive survey on routing protocols for IoTsulaiman_karim
 
WSN network architecture -Sensor Network Scenarios & Transceiver Design Consi...
WSN network architecture -Sensor Network Scenarios & Transceiver Design Consi...WSN network architecture -Sensor Network Scenarios & Transceiver Design Consi...
WSN network architecture -Sensor Network Scenarios & Transceiver Design Consi...ArunChokkalingam
 
IT8602 Mobile Communication Question Bank
IT8602 Mobile Communication Question BankIT8602 Mobile Communication Question Bank
IT8602 Mobile Communication Question Bankpkaviya
 
Mac protocols for ad hoc wireless networks
Mac protocols for ad hoc wireless networks Mac protocols for ad hoc wireless networks
Mac protocols for ad hoc wireless networks Divya Tiwari
 
Routing Protocols for Wireless Sensor Networks
Routing Protocols for Wireless Sensor NetworksRouting Protocols for Wireless Sensor Networks
Routing Protocols for Wireless Sensor NetworksDarpan Dekivadiya
 
Transmission impairments(presentation)
Transmission impairments(presentation)Transmission impairments(presentation)
Transmission impairments(presentation)Vivek Kumar
 
Physical design of io t
Physical design of io tPhysical design of io t
Physical design of io tShilpaKrishna6
 
Lecture5 teletraffic
Lecture5 teletrafficLecture5 teletraffic
Lecture5 teletrafficmazlina1202
 
FEC-Forward Error Correction for Optics Professionals
FEC-Forward Error Correction for Optics ProfessionalsFEC-Forward Error Correction for Optics Professionals
FEC-Forward Error Correction for Optics ProfessionalsMapYourTech
 
Computer Network - Network Layer
Computer Network - Network LayerComputer Network - Network Layer
Computer Network - Network LayerManoj Kumar
 

What's hot (20)

ALOHA Protocol (in detail)
ALOHA Protocol (in detail)ALOHA Protocol (in detail)
ALOHA Protocol (in detail)
 
Comprehensive survey on routing protocols for IoT
Comprehensive survey on routing protocols for IoTComprehensive survey on routing protocols for IoT
Comprehensive survey on routing protocols for IoT
 
WSN network architecture -Sensor Network Scenarios & Transceiver Design Consi...
WSN network architecture -Sensor Network Scenarios & Transceiver Design Consi...WSN network architecture -Sensor Network Scenarios & Transceiver Design Consi...
WSN network architecture -Sensor Network Scenarios & Transceiver Design Consi...
 
IT8602 Mobile Communication Question Bank
IT8602 Mobile Communication Question BankIT8602 Mobile Communication Question Bank
IT8602 Mobile Communication Question Bank
 
Mac protocols for ad hoc wireless networks
Mac protocols for ad hoc wireless networks Mac protocols for ad hoc wireless networks
Mac protocols for ad hoc wireless networks
 
Leach protocol
Leach protocolLeach protocol
Leach protocol
 
Routing Protocols for Wireless Sensor Networks
Routing Protocols for Wireless Sensor NetworksRouting Protocols for Wireless Sensor Networks
Routing Protocols for Wireless Sensor Networks
 
kpi-analysis-nsn
kpi-analysis-nsnkpi-analysis-nsn
kpi-analysis-nsn
 
Quality of Service
Quality of ServiceQuality of Service
Quality of Service
 
BIT Error Rate
BIT Error RateBIT Error Rate
BIT Error Rate
 
Gsm signaling
Gsm signalingGsm signaling
Gsm signaling
 
Transmission impairments(presentation)
Transmission impairments(presentation)Transmission impairments(presentation)
Transmission impairments(presentation)
 
GSM channels
GSM channelsGSM channels
GSM channels
 
Physical design of io t
Physical design of io tPhysical design of io t
Physical design of io t
 
Unit 4
Unit 4Unit 4
Unit 4
 
8051 timer counter
8051 timer counter8051 timer counter
8051 timer counter
 
Issues in Data Link Layer
Issues in Data Link LayerIssues in Data Link Layer
Issues in Data Link Layer
 
Lecture5 teletraffic
Lecture5 teletrafficLecture5 teletraffic
Lecture5 teletraffic
 
FEC-Forward Error Correction for Optics Professionals
FEC-Forward Error Correction for Optics ProfessionalsFEC-Forward Error Correction for Optics Professionals
FEC-Forward Error Correction for Optics Professionals
 
Computer Network - Network Layer
Computer Network - Network LayerComputer Network - Network Layer
Computer Network - Network Layer
 

Similar to Topic5

Medium Access PROTOCOL b yENGR. FAWAD KHAN UET BANNU KP PAKISTAN
Medium Access PROTOCOL b yENGR.  FAWAD KHAN UET BANNU KP PAKISTANMedium Access PROTOCOL b yENGR.  FAWAD KHAN UET BANNU KP PAKISTAN
Medium Access PROTOCOL b yENGR. FAWAD KHAN UET BANNU KP PAKISTANirfan sami
 
Mobile transport layer - traditional TCP
Mobile transport layer - traditional TCPMobile transport layer - traditional TCP
Mobile transport layer - traditional TCPVishal Tandel
 
tcp-wireless-tutorial.ppt
tcp-wireless-tutorial.ppttcp-wireless-tutorial.ppt
tcp-wireless-tutorial.pptRadwan Mahmoud
 
Lecture 19 22. transport protocol for ad-hoc
Lecture 19 22. transport protocol for ad-hoc Lecture 19 22. transport protocol for ad-hoc
Lecture 19 22. transport protocol for ad-hoc Chandra Meena
 
lec 3 4 Core Delays Thruput Net Arch.ppt
lec 3 4 Core Delays Thruput Net Arch.pptlec 3 4 Core Delays Thruput Net Arch.ppt
lec 3 4 Core Delays Thruput Net Arch.pptMahamKhurram4
 
Mediumaccesscontrol
MediumaccesscontrolMediumaccesscontrol
MediumaccesscontrolVk Sreedhar
 
Performance analysis of collision alleviating distributed coordination functi...
Performance analysis of collision alleviating distributed coordination functi...Performance analysis of collision alleviating distributed coordination functi...
Performance analysis of collision alleviating distributed coordination functi...MdAhasanulAlam
 
A preamble-based approach for Providing QOS support in Wireless Sensor Network
A preamble-based approach for Providing QOS support in Wireless Sensor NetworkA preamble-based approach for Providing QOS support in Wireless Sensor Network
A preamble-based approach for Providing QOS support in Wireless Sensor Networkdiala wedyan
 
Direct Link Lan
Direct Link LanDirect Link Lan
Direct Link Lanyanhul
 

Similar to Topic5 (20)

Medium Access PROTOCOL b yENGR. FAWAD KHAN UET BANNU KP PAKISTAN
Medium Access PROTOCOL b yENGR.  FAWAD KHAN UET BANNU KP PAKISTANMedium Access PROTOCOL b yENGR.  FAWAD KHAN UET BANNU KP PAKISTAN
Medium Access PROTOCOL b yENGR. FAWAD KHAN UET BANNU KP PAKISTAN
 
wireless.ppt
wireless.pptwireless.ppt
wireless.ppt
 
Mobile transport layer - traditional TCP
Mobile transport layer - traditional TCPMobile transport layer - traditional TCP
Mobile transport layer - traditional TCP
 
3_pstn2.ppt
3_pstn2.ppt3_pstn2.ppt
3_pstn2.ppt
 
tcp-wireless-tutorial.ppt
tcp-wireless-tutorial.ppttcp-wireless-tutorial.ppt
tcp-wireless-tutorial.ppt
 
Lecture 19 22. transport protocol for ad-hoc
Lecture 19 22. transport protocol for ad-hoc Lecture 19 22. transport protocol for ad-hoc
Lecture 19 22. transport protocol for ad-hoc
 
lec 3 4 Core Delays Thruput Net Arch.ppt
lec 3 4 Core Delays Thruput Net Arch.pptlec 3 4 Core Delays Thruput Net Arch.ppt
lec 3 4 Core Delays Thruput Net Arch.ppt
 
Link layer
Link layerLink layer
Link layer
 
Ch5 data layer network
Ch5 data layer networkCh5 data layer network
Ch5 data layer network
 
DSR,LSR,IGMP,RIP,OSPF.ppt
DSR,LSR,IGMP,RIP,OSPF.pptDSR,LSR,IGMP,RIP,OSPF.ppt
DSR,LSR,IGMP,RIP,OSPF.ppt
 
Mediumaccesscontrol
MediumaccesscontrolMediumaccesscontrol
Mediumaccesscontrol
 
Performance analysis of collision alleviating distributed coordination functi...
Performance analysis of collision alleviating distributed coordination functi...Performance analysis of collision alleviating distributed coordination functi...
Performance analysis of collision alleviating distributed coordination functi...
 
A preamble-based approach for Providing QOS support in Wireless Sensor Network
A preamble-based approach for Providing QOS support in Wireless Sensor NetworkA preamble-based approach for Providing QOS support in Wireless Sensor Network
A preamble-based approach for Providing QOS support in Wireless Sensor Network
 
Ieee 802.11.n
Ieee 802.11.nIeee 802.11.n
Ieee 802.11.n
 
Ieee 802.11.n
Ieee 802.11.nIeee 802.11.n
Ieee 802.11.n
 
Ieee 802.11.n
Ieee 802.11.nIeee 802.11.n
Ieee 802.11.n
 
Direct Link Lan
Direct Link LanDirect Link Lan
Direct Link Lan
 
UnIT VIII manet
UnIT VIII manetUnIT VIII manet
UnIT VIII manet
 
A Performance Comparison of TCP Protocols
A Performance Comparison of TCP Protocols A Performance Comparison of TCP Protocols
A Performance Comparison of TCP Protocols
 
Raj
RajRaj
Raj
 

More from hinalala

Sensys ch5-mac
Sensys ch5-macSensys ch5-mac
Sensys ch5-machinalala
 
Mac adhoc (1)
Mac adhoc (1)Mac adhoc (1)
Mac adhoc (1)hinalala
 
Lect04 (1)
Lect04 (1)Lect04 (1)
Lect04 (1)hinalala
 
Habibullah cdmabased
Habibullah cdmabasedHabibullah cdmabased
Habibullah cdmabasedhinalala
 
Fuessler wkmh03
Fuessler wkmh03Fuessler wkmh03
Fuessler wkmh03hinalala
 
Cm chou 20050124
Cm chou 20050124Cm chou 20050124
Cm chou 20050124hinalala
 

More from hinalala (12)

Wsn 08
Wsn 08Wsn 08
Wsn 08
 
View
ViewView
View
 
Tr 97 67
Tr 97 67Tr 97 67
Tr 97 67
 
Sensys ch5-mac
Sensys ch5-macSensys ch5-mac
Sensys ch5-mac
 
Mac adhoc
Mac adhocMac adhoc
Mac adhoc
 
Mac adhoc (1)
Mac adhoc (1)Mac adhoc (1)
Mac adhoc (1)
 
Mac
MacMac
Mac
 
Lect04
Lect04Lect04
Lect04
 
Lect04 (1)
Lect04 (1)Lect04 (1)
Lect04 (1)
 
Habibullah cdmabased
Habibullah cdmabasedHabibullah cdmabased
Habibullah cdmabased
 
Fuessler wkmh03
Fuessler wkmh03Fuessler wkmh03
Fuessler wkmh03
 
Cm chou 20050124
Cm chou 20050124Cm chou 20050124
Cm chou 20050124
 

Topic5

  • 1. 1 1 Medium access control in Wireless Ad-hoc Networks Presented by Jin Xu and Lan Nguyen 2 Outline Classifications of MAC protocols Five Phase Reservation Protocol (FPRP) Distributed Wireless Ordering Protocol (DWOP) Conclusions 3 Classifications of MAC protocols Contention-based A node contends with its neighbors to access the channel No QoS guarantees MACAW, FAMA, BTMA Contention-based with reservation Reserve bandwidth a priori Can provide QoS support to real time traffic D-PRMA, CATA, FPRP Contention-based with scheduling Focus on transmission scheduling of the nodes Fair and no starvation DPS, DWOP, DLPS 4 Five-Phase Reservation Protocol (FPRP) Contention-based with reservation Single channel time division multiple access Fully distributed (synchronized) Slot reservation using a 5 phase process Parallel Localized process Scalable (insensitive to network size) 5 FPRP: the model Nodes keep perfect timing A link between 2 nodes is noiseless, symmetric The network topology not change when FPRP is performed When multiple packets arrive at a node, they are destroyed A node can tell whether 0, 1, or multiple packets are transmitted when in receiving mode Every node has a unique ID 6 FPRP: overview Time is divided into frames
  • 2. 2 7 FPRP: 5 phase Reservation request (RR) Collision report (CR) Reservation confirmation (RC) Reservation acknowledgement (RA) Packing and elimination (P/E) A node keeps global timing, and knows when a 5-phase cycle starts. A node can transmit or receive, but cannot do both at the same time. 8 Phase 1: Reservation request A node which wants to make a reservation sends a Reservation Request packet (RR) with probability p Other nodes listen Requesting Node 9 Phase 2: Collision report If a node receives multiple RR’s in phase 1, it transmits a Collision Report packet (CR) o.w. silent Requesting node (RN) transmission node (TN) 10 Phase 3: Reservation confirmation TN sends Reservation Confirmation packet (RC) Every node (1 hop away) which receives RC know the slot has been reserved cease contention, receiving 11 Phase 4: Reservation acknowledgement A node ack a RC just received by sending a Reservation Ack packet (RA) Inform nodes 2 hops away Not transmitting Prevent isolated node from transmitting Resolve isolated deadlock 12 Phase 5: packing / elimination Every node 2 hops away from TN sends a Packing packet (PP) A node receiving PP learns there is a TN 3 hops away Adjust its contention prob. to reuse time slot efficiently TN sends an Elimination packet (EP) with a probability of 0.5 Attempt to resolve a non-isolated deadlock
  • 3. 3 13 An example 14 How to calculate the contention probability nc : # of nodes that contend within 2 hops nb : # of nodes within 2 hops that have to contend in next slot due to a nearby success (cannot contend in current slot) R1: a portion of 1 hop neighbors from success cease to contend in the current slot R2: 2 hop neighbors R3: 3 hop neighbors 15 How to calculate the contention probability At the beginning of a reservation slot, a node resets its nc and nb: nc = nb nb = 0 After every reservation cycle, on hearing an: Idle: nc = nc – 1 Collision: nc = nc + 1 / (e - 2) Success: 0 hop: done; 1 hop: nc = nc – 1 nb = nb + nc R1 nc =nc(1 – R1) 2 hops: nc = nc – 1 nb = nb + nc R2 nc =nc(1 – R2) 3 hops: nb = nb + nc R3 nc =nc(1 – R3) P = 1/nc 16 Simulation results The # of nodes is 100, 200, 300 and 400 from bottom to top 17 Effects of nodal mobility The observation time t is 0, 0.5, 1, 2, 4, 8, 16, 32, 64, 128 seconds from bottom to top BM RCSt = 0 18 Protocol considerations and applications Time synchronization issue: GPS can provide accurate global time Some applications: TDMA schedule produced can be used to transmit user generated packets Make reservations for network control traffic Can provide good service for multimedia (voice) traffic
  • 4. 4 19 Conclusions for FPRP Fully distributed, requiring no a prior knowledge about the network Generate transmission schedules with low amount of overhead Not affected much by the network size and nodal mobility Suitable for use in large, mobile ad hoc network 20 Problems of IEEE 802.11 Unfair channel access Due to random access nature of wireless 802.11 Hard to support QoS (priority routing/scheduling e.g. diverges significantly from FIFO order Even worse in more complex topologies, e.g. Asymmetric information Perceived collision 21 Asymmetric information Symmetric information sharing All nodes are within radio range of each other, hence can hear all the RTS and CTS All nodes have the same probability of accessing the channel Asymmetric information sharing (opposite) All nodes are not within radio range of each other Nodes have unequal channel access probabilities 22 Example Node 1 does not know about flow B, so it has to send RTS randomly Node 2 may not be able to send CTS due to either: Deferral for flow B transmission Didn’t receive RTS due to collision with flow B Node 3 knows about flow A via node 2, hence it knows the “right time” to send RTS Flow A: 5% Flow B: 95% 23 A possible solution MACAW Node 2 sends Request-for-RTS (RRTS) to node 1 Upon receiving RRTS, node 1 sends RTS immediately Works only if node 2 has received RTS Not included in IEEE 802.11 24 Perceived collision In the previous example, flow B has info about flow A (not the vice versa) -> flow B gets more channel access But this is not always the case As knowing more information about other flows makes a flow defer access to more flows
  • 5. 5 25 Perceived collision (cont.) Flows A and C can access the channel simultaneously Flow B contains information about both flows A & C, i.e. it has to wait for both flows A & C Flow A: 36% Flow B: 28% Flow C: 36% 26 DWOP protocol Goals FIFO-like behavior Piggy-backing arrival times Modifying IEEE 802.11 Receiver participation Stale entry detection 27 Goals of DWOP Providing fair channel access Scheduling packets in the order that approximates a reference scheduler (e.g. FIFO) 28 FIFO-like behavior Wireless contention for the channel access requires a scheduler for all the contending nodes For FIFO, packet priority is packet’s arrival time Contending nodes share the arrival times A scheduling table is needed 29 Piggy-backing the arrival times A sender (receiver) piggy- backs the arrival time of its current (highest priority) packet into RTS (CTS) A sender (receiver) also piggy-backs the arrival time of its next highest priority packet into the DATA (ACK) packet’s transmission 30 Piggy-backing (cont.) Nodes overhearing RTS, CTS, DATA, ACK packets add the attached arrival time into their sorted scheduling tables along with their own packets’ arrival times Also, nodes remove the arrival time entry when overhearing the completion of DATA and ACK
  • 6. 6 31 Modifying IEEE 802.11 If the channel is busy, behave like IEEE 802.11 If the channel is idle, check the scheduling table If its packet has the highest priority, send RTS Else defer as in IEEE 802.11 32 Does it work all the time ? Only if all nodes are within radio range of each other If not -> asymmetric information Node 1 is not aware of flow B -> behaves like IEEE 802.11 - > try to contend the access continuously and randomly Flow B knows the arrival time of flow A -> defers if it has lower priority -> less aggressive than flow A -> Flow B gets less bandwidth 33 Receiver participation Receiver of A (node 2) know the arrival time of flow B “warn” node 1 include an out-of-order notification in CTS/ACK Upon receiving the out- of-order notification Node 1 finishes it current transmission Node 1 goes into a backoff 34 Receiver participation (cont.) Node 1 is allowed to complete its current out-of-order transmission Only approximate FIFO is achieved DWOP can be modified to achieve perfect FIFO But tradeoff between perfect FIFO scheduling and network utilization 35 Stale entry elimination Stale arrival time entry occurs in a node if: Does not receive ACK (due to collision) Stale entry detection If there is a deletion below its packet’s position in the scheduling table Reaction Remove the first entry in the scheduling table 36 Simulation experiments Experiment setup ns-2 simulator with cmu-wireless extension Data packet size: 1000 bytes Channel bandwidth: 2Mbps CBR flows Three topologies: Asymmetric information topology Perceived collision topology More complex topology
  • 7. 7 37 Simulation experiments (cont.) Performance metrics Fair channel access FIFO-like behavior Ideal FIFO Switch to another flow after one packet transmission 38 Asymmetric information 39 Perceived collision topology 40 More complex topology 41 Summary of results DWOP has less deviation from FIFO DWOP: deviation bounded up to 4 packets IEEE 802.11: unbounded deviation DWOP achieves better fairness of accessing the channel than IEEE 802.11 42 Limitations of DWOP Flows of one single hop Fixed packet size FIFO scheduler Total number of received packets is about 1/3 less than IEEE 802.11 Does not consider Mobility Channel errors (ECF [3]) Variable packet size Other reference schedulers
  • 8. 8 43 DWOP conclusions IEEE 802.11 does not take care of fair channel access well IEEE 802.11 diverges significantly from FIFO DWOP provides fairness of channel access DWOP approximates the FIFO order better DWOP can be applied to other schedulers QoS can be integrated in DWOP 44 Reference [1]. A Five-Phase Reservation Protocol (FPRP) for Mobile Ad Hoc Networks, C Zhu, M S Corson, Proceedings of IEEE INFOCOM, 1998. [2]. Ordered Packet Scheduling in Wireless Ad Hoc Networks: Mechanisms and Performance Analysis, V. Kanodia, C. Li, A. Sabharwal, B. Sadeghi, Proceedings of ACM MOBIHOC, 2002. [3]. Fair scheduling in wireless ad-hoc networks of location dependent channel errors, Chen J., Somani A. K., Proceedings of the 2003 IEEE International, 2003. 45 Question ?