jennifer c. hou department of computer science university of illinois at urbana-champaign
DESCRIPTION
DAWN: Dynamic Ad-hoc Wireless Networks Progress Report Presentation. Jennifer C. Hou Department of Computer Science University of Illinois at Urbana-Champaign September 10, 2014. Energy Efficient Network Track Power, CS Threshold, and Rate Control. PHY/MAC Control Knobs. - PowerPoint PPT PresentationTRANSCRIPT
Jennifer C. HouJennifer C. HouDepartment of Computer ScienceDepartment of Computer Science
University of Illinois at Urbana-ChampaignUniversity of Illinois at Urbana-ChampaignApril 24, 2023April 24, 2023
DAWN: Dynamic Ad-hoc Wireless NetworksProgress Report Presentation
Energy Efficient Network TrackPower, CS Threshold, and Rate Control
DEPARTMENT OF COMPUTER SCIENCE UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGNJennifer Hou Jennifer Hou
PHY/MAC Control KnobsTo mitigate interference and maximize the network capacity, there are several control knobs:
Transmit power power/topology controlCarrier sense threshold trade-off between spatial reuse and interference levelSpatial diversity scheduling consecutive transmission for interference-free connectionsChannel diversity use of non-overlapping channels
DEPARTMENT OF COMPUTER SCIENCE UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGNJennifer Hou Jennifer Hou
Power Control
Definition: Each node adjusts its transmission power so as to maintain network connectivity using the minimum possible power.
Objectives: Maintaining network connectivity Reducing energy consumption Mitigating MAC interference Achieving network capacity and spatial reuse
DEPARTMENT OF COMPUTER SCIENCE UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGNJennifer Hou Jennifer Hou
Preliminary Work on Power Control
Local minimum spanning tree (LMST) [INFOCOM’03, IEEE TWC, IEEE TPDS]Localized algorithm Relies only on local informationPreserves connectivity.Ensures bi-directional links.Handles node heterogeneity (i.e., nodes have different maximal transmit powers)Bounds the degree of any node by 6.
DEPARTMENT OF COMPUTER SCIENCE UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGNJennifer Hou Jennifer Hou
Xi : location of node iRi : transmission range of node iLink (ji) exists if Rj | Xi – Xj |:Transmission power of node i
Total power:k-connectivity: requires to remove at least k nodes to disconnect the networkCritical total power Wc: minimum total power W for maintaining k-connectivity
i iW R
i ii iW W R
1
1
Ri
RjXj
Xi
Preliminary Work on Total PowerRequired for K-Connectivity
Poisson point process with density n in a unit-
area square
DEPARTMENT OF COMPUTER SCIENCE UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGNJennifer Hou Jennifer Hou
Question To AskIn what order does the critical total power Wc increase/decrease as the node density increases?All nodes choose common power
[Gupta & Kumar 98] studied the critical transmission range rn for 1-connectivity
[Wan & Yi 04] for k-connectivityAll nodes choose different power
[Blough 02] critical total power for 1-connectivity
Based on the total weight of minimum spanning tree
Our study: critical total power or k-connectivity
log , where ,as nn n
nr nn
DEPARTMENT OF COMPUTER SCIENCE UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGNJennifer Hou Jennifer Hou
Major ResultsMain theory: [Infocom 2005, ACM/IEEE ToN 2006]
The critical total power for maintaining k-connectivity
is with probability approaching 1
Comparison with common powerThe critical total power for k-connectivity with common power is
Allowing power control at each node reduces the total power by a factor of
1 / 2( / 2 )( )( 1)!
k nk
1 / 2 / 2( (log ) )n n
/ 2((log ) )n
DEPARTMENT OF COMPUTER SCIENCE UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGNJennifer Hou Jennifer Hou
Rescaling to Expanded Networks
Expanded networks Node density fixed
Side length L
Expected number of nodes n= L2
Allowing power controlAverage power is bounded
Using common power The common power grows as
(1)
/ 2((log ) )n
1
1ijd
L
L 'ijd
ijd
'ij ijd L d
DEPARTMENT OF COMPUTER SCIENCE UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGNJennifer Hou Jennifer Hou
Cross Layer Aspects of Power Control
Physical Layer
MAC Layer
Network Layer
Power C
ontrol Incorporating Physical Layer Characteristics
Cross Layer Design
Effect of MAC-Layer Interference
Dynamic Topology Control w.r.t.
Network TrafficNetwork Capacity
Network Lifetime
Critical Power
Analysis
Physical Layer
Incorporating Physical Layer Characteristics
DEPARTMENT OF COMPUTER SCIENCE UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGNJennifer Hou Jennifer Hou
When Power Control Meets SINR
All the topology control algorithms in literature defined the neighbor relation based on the protocol model
A link exists between nodes i and j if dij <= dmax.
TC: L(n) T(n)The protocol model ignores the effect of SINR.
What is more appropriate to define a link is the use of physical model
A link exists between i and j if
TC: L(n) x C(n) T(n) Set of node locationsSet of configurations (, max/min transmit power)
Topology that defines the neighbor relations
DEPARTMENT OF COMPUTER SCIENCE UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGNJennifer Hou Jennifer Hou
When Power Control Meets SINR
All existing topology control algorithms fail (i.e., cannot maintain network connectivity) under the physical model.We are re-investigating topology control under the physical model.
DEPARTMENT OF COMPUTER SCIENCE UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGNJennifer Hou Jennifer Hou
Controlling Carrier Sense Threshold
• The contending area can also be adapted through tuning the carrier-sensing threshold
AB C D
distance
Sign
al S
tren
gth
CS Threshold
E F
DEPARTMENT OF COMPUTER SCIENCE UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGNJennifer Hou Jennifer Hou
A large CS threshold leads tosmaller contending areaLess contending nodes within the contending areaMore concurrent transmissionHigher interference
Transmission rate depends on Signal-to-Interference-Noise Ratio
Controlling Carrier Sense Threshold
AB C D
distance
Sign
al S
tren
gth
CS Threshold
E F
DEPARTMENT OF COMPUTER SCIENCE UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGNJennifer Hou Jennifer Hou
Tradeoff AnalysisSpatial reuse can be achieved at the cost of higher interference level and lower transmission rate
High rate links Low rate links
What is the optimal CS Threshold? How does it relate to the transmit power?
DEPARTMENT OF COMPUTER SCIENCE UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGNJennifer Hou Jennifer Hou
Network CapacityNetwork capacity as a function of transmit power and carrier sense threshold [ACM Mobicom 2006]
DEPARTMENT OF COMPUTER SCIENCE UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGNJennifer Hou Jennifer Hou
Power Control vs. Data Rate
RX1r2SINR :d1
RX2r1SINR :d2
Power: PCS Threshold:
Tx1
Tx2Rx
1
Rx2
d1r2
r1
d2D1 D
2
SINR requirements
Power: PCS Threshold:
DR[2]thrSINR
DR[3]thrSINR
DR[1]thrSINR
DEPARTMENT OF COMPUTER SCIENCE UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGNJennifer Hou Jennifer Hou
Power Control vs Data Rate
Power: PCS Threshold:
'Power: P (>P)CS Threshold:
'
RX1P r2SINR : *P d1
Tx1
Tx2Rx
1
Rx2
d1r2
r1
d2D1
D2
RX2 '
P r1SINR : *P d2
SINR requirements
DR[2]thrSINR
DR[3]thrSINR
DR[1]thrSINR
DEPARTMENT OF COMPUTER SCIENCE UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGNJennifer Hou Jennifer Hou
Power and Rate ControlPRC algorithm:
A localized algorithm that enables each transmitter to adapt to the interference level that it perceives and determines its transmit power.The transmit power is so determined that the transmitter can sustain the highest possible data rate, while keeping the adverse interference effect on the other neighboring concurrent transmissions minimal.
DEPARTMENT OF COMPUTER SCIENCE UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGNJennifer Hou Jennifer Hou
Simulation SetupModified ns-2 Ver. 2.28
The interference perceived at a receiver is the collective aggregate interference from all the concurrent transmissionsEach node uses physical carrier sense to determine if the medium is freeIEEE 802.11a radios supporting 8 discrete data rate (6 ~ 54 Mbps)
Random topology3, 10, 20, 30, and 50 transmitter-receiver pairs are randomly generated in a 300m X 300m area, and represent sparsely, moderately, and densely populated networks, respectively,.
Algorithms used for evaluationsStaticGreedy Power Control (GPC)Power and Rate Control (PRC)
DEPARTMENT OF COMPUTER SCIENCE UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGNJennifer Hou Jennifer Hou
Simulation Results
Performance gain mainly because ofHigher concurrent transmissions
Simulation Track1. Expediting Wireless Simulation2. Incorporating Model Checking into Simulation
DEPARTMENT OF COMPUTER SCIENCE UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGNJennifer Hou Jennifer Hou
Expediting Wireless Simulation
Our profiling work indicates more than 50% of the execution time is spent on event scheduling and channel related activity handling.Can we expedite simulation by reducing the number of unnecessary events while not impairing the accuracy. Proportion of the execution time that is
spent on event enqueueing in a 100-node ad hoc network over a 1000mX1000m field. There are 40 CBR connections carrying a total of 120 packets/sec. Traffic (pkt. Size = 512B)
[1] Chunyu Hu and Jennifer C. Hou, ``A reactive channel model for expediting wireless network simulation,'' ACM SIGMETRICS, Banff, Alberta, Canada, June 2005
DEPARTMENT OF COMPUTER SCIENCE UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGNJennifer Hou Jennifer Hou
Reactive Channel Model
R
L
The channel only notifies nodes in the following sets of the signal-arrival event
Nodes in range RNodes in (range L but not R) that are registered
When does a node register?
Whenever it needs to monitor the channel status, e.g., when it would like to gain access to the channel or when it is in the process of receiving a signal
DEPARTMENT OF COMPUTER SCIENCE UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGNJennifer Hou Jennifer Hou
A Case Study: IEEE 802.11 MAC
State To register?
transmitting * ☒receiving ☑
idlebacking-off ☑
deferring ☑
-- ☒sleep ☒
turn-off ☒
*: assuming half-duplex radio is used
DEPARTMENT OF COMPUTER SCIENCE UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGNJennifer Hou Jennifer Hou
RCM Performance (Execution Time) RCM
DEPARTMENT OF COMPUTER SCIENCE UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGNJennifer Hou Jennifer Hou
RCM Performance(Memory Consumption)
DEPARTMENT OF COMPUTER SCIENCE UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGNJennifer Hou Jennifer Hou
Virtual Wireless Ad-Hoc Network (J-Sim)Champaign-Urbana Wireless
Community Network(Currently 40 wireless nodesin downtown Urbana; expectedto extend to 100 nodes providingfull coverage of Champaign and Urbana).
Integration of Real/Virtual Integration of Real/Virtual WorldsWorlds
• Channel behavior modeling• Physical capacity analysis• Incentive-based resource management• Multi-radio, multi-path routing• Cross-layer optimization
DEPARTMENT OF COMPUTER SCIENCE UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGNJennifer Hou Jennifer Hou
Marriage of Modeling Checking and SimulationMarriage of Modeling Checking and Simulation
s0
s1
s2
s3
s4
s5
s6
s7
X
J-Sim
[1] Ahmed Sobeih, Mahesh Viswanathan and Jennifer C. Hou, “Check and Simulate: A Case for Incorporating Model Checking in Network Simulation,” Proceedings of the ACM-IEEE International Conference on Formal Methods and Models for Codesign (ACM-IEEE MEMOCODE 2005), San Diego, CA, June 2005.
s0
s1
s2
s3
s4
s5
s6
s7
X
J-Sim w/ MC
• Stateful on-the-fly explicit-state model checking into J-Sim• Explore the state space of a network protocol up to a (configurable) maximum depth of transitions• No changes to the core design and implementation of J-Sim
DEPARTMENT OF COMPUTER SCIENCE UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGNJennifer Hou Jennifer Hou
An Overview of Our WorkAn Overview of Our Work• Build the model checker as a component in J-Sim
P1 P2 Pn
Model Checker
J-Sim
Error Trace / No Error
Initial State
Current State
Next State
Component
Port
Communication via ports
DEPARTMENT OF COMPUTER SCIENCE UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGNJennifer Hou Jennifer Hou
Evaluation and ResultsEvaluation and Results
• Two case studies: AODV and Directed Diffusion• Representative routing and data dissemination protocols• Reasonably complex network protocols
• 1200 – 1400 LOC (excluding the J-Sim library)• Safety property:
• The loop-free property of routing/data dissemination paths
[2] Ahmed Sobeih, Mahesh Viswanathan, Darko Marinov and Jennifer C. Hou, “Finding Bugs in Network Protocols Using Simulation Code and Protocol-Specific Heuristics,” Proceedings of the International Conference on Formal Engineering Methods (ICFEM 2005), Springer-Verlag LNCS 3785, Manchester, United Kingdom, November 2005.
• Summary of our discoveries: • A previously unknown bug in the J-Sim implementation of AODV• A previously unknown deficiency in directed diffusion
DEPARTMENT OF COMPUTER SCIENCE UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGNJennifer Hou Jennifer Hou
Guiding Model Checking with Network PropertiesWe have developed search heuristics that exploit properties inherent to the network protocol and the safety property being checked and better guide the model checker to discover counter examples. An interesting and important research question is how to determine a suitable BeFS strategy for a specific network protocol.