Impact Analysis of Different Scheduling and Retransmission Techniques on an Underwater Routing Protocol

Impact Analysis of Different Scheduling and Retransmission Techniques on an Underwater Routing Protocol Salvador Climent 1 Nirvana Meratnia 2 Juan Vicente Capella 1 1 Universitat Politècnica de València 2 University of Twente Workshop on UnderWater Networks, 2011 Climent, Meratnia, Capella (UPV and UT) WUWNet 2011 1 / 34

Outline 1 Underwater Transmission Problems 2 Contributions 3 Routing protocol: EDETA 4 Scheduling and Retransmission Techniques Combination of scheduling and retransmission techniques used 5 Performance evaluation Simulation configuration Simulation results 6 Conclusions and Future work Climent, Meratnia, Capella (UPV and UT) WUWNet 2011 2 / 34

Underwater Transmission Problems Differences between WSNs and UWSNs make current protocols developed for WSNs unsuitable or inefficient for UWSNs. Climent, Meratnia, Capella (UPV and UT) WUWNet 2011 4 / 34

Underwater Transmission Problems Different signal attenuation depending on distance and frequency. Signal spreading proportional to the distance. Long propagation delay. Climent, Meratnia, Capella (UPV and UT) WUWNet 2011 5 / 34

Underwater Transmission Problems Given the long propagation delay: TDMA and CSMA might not perform well. An schedule algorithm can avoid collisions and overlap transmissions. Climent, Meratnia, Capella (UPV and UT) WUWNet 2011 6 / 34

Contributions Analyze performance of different delay-aware and non-delay aware scheduling techniques. Combine them with simple retransmission techniques. The analysis is done using a routing protocol named EDETA. Climent, Meratnia, Capella (UPV and UT) WUWNet 2011 8 / 34

EDETA Hierarchical protocol Nodes arrange themselves in clusters. The cluster-heads (CHs) are know a priori. CHs arrange themselves in a tree structure with the sink as root. Climent, Meratnia, Capella (UPV and UT) WUWNet 2011 10 / 34

EDETA Corporative network Sink node Clusterhead node Normal node Climent, Meratnia, Capella (UPV and UT) WUWNet 2011 11 / 34

Scheduling advantages Scheduling allows to avoid collisions and can overlap transmissions which can reduce the energy consumption and the delays. There should be an extra time for retransmission in case of packet errors. Climent, Meratnia, Capella (UPV and UT) WUWNet 2011 13 / 34

Delay-aware schedule The delay between a node and its CH is estimated at the configuration phase. The delay between a CH and its parent is estimated at the configuration phase. The schedule is done taking into account these delays. Climent, Meratnia, Capella (UPV and UT) WUWNet 2011 14 / 34

Delay-aware schedule Based on scheduling constrains from: W. A. P. van Kleunen et al. MAC Scheduling in Large-scale Underwater Acoustic Networks. 8th International Joint Conference on e-business and Telecommunication, ICETE 2011. W. A. P. van Kleunen et al. A Set of Simplified Scheduling Constraints for Underwater Acoustic MAC Scheduling. The 3rd International Workshop on Underwater Networks, WUnderNet-2011. Climent, Meratnia, Capella (UPV and UT) WUWNet 2011 15 / 34

TAck TDMA schedule with acknowledgement. Slots have to include the maximum propagation time. Each transmission is scheduled two times. Main Tx Backup Tx Data Ack Data Ack Tx Prop Tx Prop Tx Prop Tx Prop Climent, Meratnia, Capella (UPV and UT) WUWNet 2011 16 / 34

TnoAck TDMA schedule without acknowledgement. Slots have to include the maximum propagation time. There is no backup transmission. Data Data Tx Prop Tx Prop Climent, Meratnia, Capella (UPV and UT) WUWNet 2011 17 / 34

DAck Delay-aware schedule with acknowledgement. Transmissions can be overlapped. Each transmission is scheduled two times. DATA ACK DATA ACK DATA ACK DATA ACK DATA ACK DATA ACK DATA Avoid interference with ACK at sender Time Climent, Meratnia, Capella (UPV and UT) WUWNet 2011 18 / 34

DnoAck Delay-aware schedule without acknowledgement. Transmissions can be overlapped. There is no backup transmission. Receiver Time Climent, Meratnia, Capella (UPV and UT) WUWNet 2011 19 / 34

DnoAck2 Delay-aware schedule without acknowledgement Transmissions can be overlapped. The data is sent twice. DATA DATA DATA DATA DATA DATA DATA DATA Time Climent, Meratnia, Capella (UPV and UT) WUWNet 2011 20 / 34

Simulator ns-3 simulator. UAN module for the underwater layer. Climent, Meratnia, Capella (UPV and UT) WUWNet 2011 22 / 34

Scenario configuration Three different deployment areas: 100 100 150 150 200 200 Three different node densities: 50 nodes 100 nodes 200 nodes Climent, Meratnia, Capella (UPV and UT) WUWNet 2011 23 / 34

Radio configuration Transmission range of 100 meters. Transmission speed 500 bps. Data sent periodically every 250 seconds. Nodes start with 150 Joules. Tx: 0.203 watts Rx and Idle: 0.024 watts Sleep: 3 10 6 watts Climent, Meratnia, Capella (UPV and UT) WUWNet 2011 24 / 34

TnoAck vs DnoAck. Delay Delay heavily influenced by the network topology. Short transmission range: 100 meters Propagation delay is not the dominant factor. 2.5 2 Packet delay (s) 1.5 1 0.5 0 50n100m 50n150m 50n200m 100n100m 100n150m Scenarios 100n200m 200n100m 200n150m 200n200m DnoAck TnoAck Climent, Meratnia, Capella (UPV and UT) WUWNet 2011 25 / 34

TnoAck vs DnoAck. Energy Same energy consumption for the leaf nodes CHs using delay-aware schedule have less energy consumption. Energy (J) 105 100 95 90 85 80 50n100m 50n150m 50n200m 100n100m 100n150m Scenario 100n200m 200n100m 200n150m 200n200m DnoAck TnoAck (a) Leaf nodes energy consumption Energy (J) 300 250 200 150 100 50 0 50n100m 50n150m 50n200m 100n100m 100n150m Scenario 100n200m 200n100m 200n150m 200n200m DnoAck TnoAck (b) CH nodes energy consumption Climent, Meratnia, Capella (UPV and UT) WUWNet 2011 26 / 34

TAck vs DAck. Delay TDMA delay increases when new control packets are introduced. The delay-aware schedule can optimize and reduce these delays. 14 12 Packet delay (s) 10 8 6 4 2 0 50n100m 50n150m 50n200m 100n100m 100n150m 100n200m Scenarios 200n100m 200n150m 200n200m DAck TAck Climent, Meratnia, Capella (UPV and UT) WUWNet 2011 27 / 34

TAck vs DAck. Energy CHs using delay-aware schedule have less energy consumption. Leaf nodes spend the same time awake in the two alternatives. Energy (J) 110 105 100 95 90 85 80 50n100m 50n150m 50n200m 100n100m 100n150m 100n200m 200n100m 200n150m 200n200m Scenarios (c) Leaf nodes energy consumption DAck TAck Energy (J) 500 450 400 350 300 250 200 150 100 50 0 50n100m 50n150m 50n200m 100n100m 100n150m 100n200m 200n100m 200n150m 200n200m Scenarios (d) CH nodes energy consumption DAck TAck Climent, Meratnia, Capella (UPV and UT) WUWNet 2011 28 / 34

DAck vs DnoAck2. Delay As expected DnoAck2 has less delay since nodes don t wait for the ACK. Packet delay (s) 9 8 7 6 5 4 3 2 1 0 50n100m 50n150m 50n200m 100n100m 100n150m Scenarios 100n200m 200n100m 200n150m 200n200m DnoAck2 DAck Climent, Meratnia, Capella (UPV and UT) WUWNet 2011 29 / 34

DAck vs DnoAck2. Energy Leaf nodes with DnoAck2 alternative have an overhead. CH nodes with DnoAck2 have the lowest energy consumption. Energy (J) 120 100 80 60 40 20 0 50n100m 50n150m 50n200m 100n100m 100n150m Scenario 100n200m 200n100m 200n150m 200n200m DnoAck2 DAck (e) Leaf nodes energy consumption Energy (J) 450 400 350 300 250 200 150 100 50 0 50n100m 50n150m 50n200m 100n100m 100n150m Scenario 100n200m 200n100m 200n150m 200n200m DnoAck2 DAck (f) CH nodes energy consumption Climent, Meratnia, Capella (UPV and UT) WUWNet 2011 30 / 34

Conclusions The differences between TnoAck and DnoAck are given by the configuration phase. CHs using DAck has less energy consumption and there is lower delay. With the DnoAck2 alternative leaf nodes have the highest energy consumption. Climent, Meratnia, Capella (UPV and UT) WUWNet 2011 32 / 34

Future work Include different topologies. Guidelines on the optimal use of retransmission techniques and different scheduling techniques under different circumstances. Climent, Meratnia, Capella (UPV and UT) WUWNet 2011 33 / 34