Scheduling Granularity in Underwater Acoustic Networks Kurtis Kredo - - PowerPoint PPT Presentation

Scheduling Granularity in Underwater Acoustic Networks 1/21

Scheduling Granularity in Underwater Acoustic Networks

Kurtis Kredo II1 Prasant Mohapatra2

1Electrical and Computer Engineering Department

California State University, Chico

2Computer Science Department

University of California, Davis

ACM WUWNet December 1, 2011

Scheduling Granularity in Underwater Acoustic Networks 2/21 Introduction

Underwater Scheduling Spectrum

Goals Explore the many scheduling options in underwater networks Find balance between complexity and performance Tight Scheduling Advantages Decreased frame size

Higher throughput Lower latency

Increased sleep opportunities Tight Scheduling Disadvantages Higher state overhead Inflexible communication

• pportunities

Less resiliency to synchronization error

Scheduling Granularity in Underwater Acoustic Networks 3/21 Channel Scheduling Spectrum

Frame Size CDF

Underwater Network

0.2 0.4 0.6 0.8 1 100 200 300 400 500 600 700 800 Cumulative Distribution Frame Size (slots) TDMA Node Group Link Slot

Terrestrial Network

0.2 0.4 0.6 0.8 1 150 200 250 300 350 400 450 500 Cumulative Distribution Frame Size (slots) TDMA Node Group Link Slot

Scheduling Granularity in Underwater Acoustic Networks 4/21 Channel Scheduling Spectrum Five Scheduling Methods

Scheduling Spectrum Examples

Network

B A D C E

A→C: 4 Slots B→D: 2 Slots Example Frames

B C D A B C D A TDMA Node RX Intf TX

Scheduling Granularity in Underwater Acoustic Networks 5/21 Channel Scheduling Spectrum Five Scheduling Methods

Scheduling Spectrum Examples

Network

B A D C E

A→C: 4 Slots B→D: 2 Slots Example Frames

TX Intf RX B C D B C D A A Node Group

Scheduling Granularity in Underwater Acoustic Networks 6/21 Channel Scheduling Spectrum Five Scheduling Methods

Scheduling Spectrum Examples

Network

B A D C E

A→C: 4 Slots B→D: 2 Slots Example Frames

B C D A B C D A Group Link TX Intf RX

Scheduling Granularity in Underwater Acoustic Networks 7/21 Channel Scheduling Spectrum Five Scheduling Methods

Scheduling Spectrum Examples

Network

B A D C E

A→C: 4 Slots B→D: 2 Slots Example Frames

B C D A Link B C D A Slot TX Intf RX

Scheduling Granularity in Underwater Acoustic Networks 8/21 Channel Scheduling Spectrum Schedule State Requirements

Scheduling State Requirements

Scheduling Options State Class Characteristics S D P TDMA Fixed-length blocks × Node Variable-length blocks × × × Group Group neighbors into rings × × × Link Single link transmission × × × Slot Multiple link transmissions × × × State Variables State Description S Slot for each element D Duration for each element P Propagation delay or ring number

Scheduling Granularity in Underwater Acoustic Networks 9/21 Scheduling Problem

Communication Characteristics

Device Capabilities Single half-duplex radio Optional DSSS with single packet reception Stationary nodes Network Conflicts

j j i j i i

RX−RX TX−TX TX−RX

k i k

TX−RX−TX Each network conflict results in a schedule constraint

Scheduling Granularity in Underwater Acoustic Networks 10/21 Scheduling Problem

TX-RX Constraint Formulation

Example Frame

sj pj

j

si ∆i s +m

Network Conflicts

j j i j i i RX−RX TX−TX TX−RX k i k TX−RX−TX

Scheduling Granularity in Underwater Acoustic Networks 10/21 Scheduling Problem

TX-RX Constraint Formulation

Example Frame

sj pj

j

si ∆i s +m

Schedule Constraints si ≥ sj + ∆j + pj T sj + pj T + m ≥ si + ∆i

Scheduling Granularity in Underwater Acoustic Networks 11/21 Scheduling Problem

TX-RX Schedule Constraint

Node j Transmits First si ≥ sj + ∆j + pj T sj + pj T + m ≥ si + ∆i ⊕ ⇓ Node j Transmits Second sj + pj T ≥ si + ∆i si + m ≥ sj + ∆j + pj T Combined Constraints si + moij ≥ sj + ∆j + pj T sj + pj T + m (1 − oij) ≥ si + ∆i ⇓ ∆i − pj T − m ≤ sj − si − moij ≤ −∆j + pj T

Scheduling Granularity in Underwater Acoustic Networks 12/21 Scheduling Problem

Scheduling Problem

Given: Propagation delays p Required transmission durations ∆ Find: Transmission slots s Satisfy all schedule constraints Bij − m ≤ sj − si − moij ≤ Bij

Scheduling Granularity in Underwater Acoustic Networks 13/21 Numerical Evaluation

Methodology

Summary Find schedules using CPLEX 13 nodes distributed in a grid 10 time slots from each node 1 time slot to each node Central sink as destination 100 random networks Evaluate:

Impact of synchronization error Use of DSSS Best scheduling options to evaluate in simulation

Topology

Scheduling Granularity in Underwater Acoustic Networks 14/21 Numerical Evaluation Throughput and Latency

Network Performance

Uplink Latency

200 400 600 800 1000 1200 1400 1600 0.1 0.2 0.3 0.4 0.5 Average Maximum Uplink Latency (Slots) Synchronization Error, σ (s) TDMA Node Group Link Distributed Scheduling Optimal Latnecy

Aggregate Throughput

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.1 0.2 0.3 0.4 0.5 Normalized Throughput Synchronization Error, σ (s) Optimal Aloha TDMA Node Group Link Slot

Summary

Performance along spectrum as expected Scheduling options scale well with synchronization error Slot provides no throughput improvement Higher throughput than optimal Aloha, without collisions

Scheduling Granularity in Underwater Acoustic Networks 15/21 Numerical Evaluation Performance with DSSS

DSSS

Aggregate Throughput

0.2 0.4 0.6 0.8 1 1.2 No DSSS 1 2 4 8 Normalized Throughput DSSS Spreading Factor TDMA Node Group Link Slot Distributed Scheduling Optimal Frame Size ST-MAC

Summary

Performance along spectrum as expected Slot provides no throughput improvement DSSS provides no benefit above SF = 1 Comparable throughput performance with ST-MAC (centralized)

Scheduling Granularity in Underwater Acoustic Networks 16/21 Simulation Results

Simulation Results

Summary Using OMNeT++ Discrete Event Simulator Topologies of 13 and 29 nodes Central sink as destination Variable data rate and transmit power 30 random networks Evaluate:

Metrics unavailable in numerical evaluation Performance compared to related protocols Protocol convergence time

Scheduling Granularity in Underwater Acoustic Networks 17/21 Simulation Results

Efficiency

Efficiency

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 20 40 60 80 100 120 Protocol Efficiency (bits/mJ) Average Node Data Rate (bits/s) Node Group Link Aloha ST-Lohi UT-Lohi

Summary Link scheduling delivers data for lower energy Link scheduling provides 300% efficiency improvement

Scheduling Granularity in Underwater Acoustic Networks 18/21 Simulation Results

Throughput and Latency

Throughput

200 400 600 800 1000 1200 1400 50 100 150 200 Network Throughput (bits/s) Average Node Data Rate (bits/s) Node Group Link Aloha ST-Lohi cUT-Lohi aUT-Lohi

Uplink Latency

500 1000 1500 2000 2500 3000 20 40 60 80 100 120 Packet Latency (s) Average Node Data Rate (bits/s) Node Group Link Aloha ST-Lohi cUT-Lohi aUT-Lohi

Summary Link scheduling has 200%–300% higher throughput Link scheduling provides low latency

Scheduling Granularity in Underwater Acoustic Networks 19/21 Simulation Results

DSSS

Link Scheduling Average Frame Size

Spreading Factor, SF Average Frame Size (slots) No DSSS 128.2 1 86.1 2 163.5 4 314.1 8 610.8

Schedule State Size (bits)

No DSSS DSSS Scheduling Method Mean Max Mean Max Node 364.9 848 165.0 464 Group 378.0 1040 178.7 656 Link 347.8 1176 146.5 400

Summary Frame size scales with SF DSSS reduces distributed schedule state

Scheduling Granularity in Underwater Acoustic Networks 20/21 Simulation Results

Scheduling Convergence

Epoch Size in Frames No DSSS DSSS Scheduling Method Mean Max Mean Max Node 17.4 22 8.2 12 Group 15.2 19 8.4 13 Link 12.2 17 7.6 10 Summary More specific scheduling converges faster Using DSSS causes faster convergence Results provide indication of required epoch size

Scheduling Granularity in Underwater Acoustic Networks 21/21 Conclusion

Conclusions

Several options for schedule granularity Numerical results evaluated five scheduling methods

TDMA scheduling provides poor performance Slot scheduling requires significant resources

Simulation results evaluated three best candidates

Link scheduling provides best balance Better performance than other methods and protocols

DSSS

Provides no benefit to traffic metrics Reduces control state requirements Decreases schedule convergence time

Scheduling Granularity in Underwater Acoustic Networks 22/21 Conclusion

Thank You!

Questions?

Scheduling Granularity in Underwater Acoustic Networks 18/21 Appendix

TDMA Constraints

Time Slot Set Size Λ = max

a

• ∆a +

pa T

• Constraint

Λ − m ≤ sb − sa − moab ≤ −Λ

Scheduling Granularity in Underwater Acoustic Networks 19/21 Appendix

Node Constraints

Time Slot Set Size Λa = ∆a + pa T Constraint Λa − m ≤ sb − sa − moab ≤ −Λb

Scheduling Granularity in Underwater Acoustic Networks 20/21 Appendix

Group Constraints

Ring Definition ra,b = p(a,b) T

• TX-TX Constraints

∆a,r− − m ≤ sa,r − sa,r− − moar−,ar ≤ ∆a,r− − m Other Constraints ∆a + min

• rI

a − rb, ra,b

• + 1 − m

≤ sb − sa − moa,b ≤ − ∆b + min

• rI

a, ra,b

• − 2rb − 1

Scheduling Granularity in Underwater Acoustic Networks 21/21 Appendix

Link and Slot Constraints

TX-TX Constraints ∆i − m ≤ sj − si − moij ≤ −∆j TX-RX Constraints ∆i − pj T − m ≤ sj − si − moij ≤ −∆j − pj T RX-RX-Constraints ∆i + pi − pj T − m ≤ sj − si − moij ≤ pi − pj T − ∆j TX-RX-TX Constraints ∆i + pi − p(js,id) T − m ≤ sj − si − moij ≤ pi − p(js,id) T − ∆j