Performance Evaluation of Route Failure Detection in Mobile Ad Hoc Networks Dimitri Marandin 4. Würzburger Workshop "IP Netzmanagement, IP Netzplanung und Optimierung" 27.-28. July 2004 www.ifn.et.tu-dresden.de/tk/ 1
Contents Introduction Reasons of route failures Reaction on route failures Route breakage detection Hello messages IEEE 802.11 MAC Feedback Passive ACKs Performance evaluation Conclusions www.ifn.et.tu-dresden.de/tk/ 2
Properties of Ad hoc Network Lack of a centralized entity Rapid mobile host movements Communication over the wireless medium Limited wireless bandwidth Limited battery power Multi-hop routing www.ifn.et.tu-dresden.de/tk/ 3
Reasons of Route Failures Node mobility Environment conditions Node failure - battery usage Hard medium contention www.ifn.et.tu-dresden.de/tk/ 4
Route Maintenance Link breakage -> Route maintenance Link breakage detected by: link-layer ACK, AODV Hello messages attempt local repair - RREQ for destination from intermediate node Route Error (RERR) message - Contains list of unreachable destinations - Sent to precursors : neighbors who recently sent packet over broken link - Propagated recursively www.ifn.et.tu-dresden.de/tk/ 5
Impact Route Failures on TCP route failure route failure detected route found TCP times out no throughput Packet discarded www.ifn.et.tu-dresden.de/tk/ 6
Optimisation Goals Correct route failure detection - avoiding false decisions Quick route failure detection Route failure detection with minimum overhead (energy, bandwidth) www.ifn.et.tu-dresden.de/tk/ 7
Link Breakage Detection Link level ACKs Hello messages (beacons) passive ACKs -used in DSR End-to-end ACKs involvement of transport layer www.ifn.et.tu-dresden.de/tk/ 8
Hello Messages Each node must periodically send broadcast packets Each receiving node must also process them Hello messages cause a large amount of control overhead Current implementations use periodic hello messages to detect local link connectivity www.ifn.et.tu-dresden.de/tk/ 9
Passive ACKs Listen for data to be forwarded Passive ack: - a node relays the packet to its neighbours via radio transmission - previous node hears this transmission and uses it as an ack All nodes are overhearing (promiscuous) No energy saving A B C Passive ack www.ifn.et.tu-dresden.de/tk/ 10
Link Layer Detection Unicast packet is successfully transmitted Route failure Based on IEEE 802.11 MAC Problem: Access link layer feedback information in off the shelf hardware Feedback Application TCP,UDP AODV. DSR,.. MAC 802.11 Phy 802.11 www.ifn.et.tu-dresden.de/tk/ 11
802.11 DCF Operation RTS B A S R Data RTS Data RTS CTS Data ACK CTS CTS C ACK Atomic unit ACK RTS CTS Data ACK www.ifn.et.tu-dresden.de/tk/ 12
IEEE 802.11 MAC Feedback After 7 failed RTS retransmissions -> packet loss -> link is failed -> route is failed -> initiating route request No additional overhead Unpredicted detection delay Incorrect detection due to channel contention www.ifn.et.tu-dresden.de/tk/ 13
Simulation Parameters Simulation Parameters (ns-2) 50 nodes 5, 10, 15, 20, 25, 30, 35 cbr connections (for each 5 runs) Static network, mobile network - max speed 20 m/s Packet size 512 bytes Simulation time 900s (200s war-up) Confidence interval 95% Performance metrics Goodput - number of unique correctly received packets Energy efficiency - total consumed energy of all nodes to total number of successfully received packets www.ifn.et.tu-dresden.de/tk/ 14
Energy Model Simplified model Decrease the energy level of the node for every packet transmitted and every packet received rxpower - receiving power 0.1 Watts txpower - transmitting power 0.3 Watts Initial energy 1000 Joules www.ifn.et.tu-dresden.de/tk/ 15
Simulations Results Overall received packets Overall recieved packets 3000 2500 2000 1500 1000 500 0 Goodput MAC feedback Hello 5 10 15 20 25 30 35 Number of connections static network www.ifn.et.tu-dresden.de/tk/ 16
Simulations Results Energy efficiency Energy, Joules 0,1 0,08 0,06 0,04 0,02 0 Hello MAC Feedback 5 10 15 20 25 30 35 Number of connections static network www.ifn.et.tu-dresden.de/tk/ 17
Simulations Results Goodput Overall received packets 3000 2500 2000 1500 1000 500 0 MAC Feedback Hello 5 10 15 20 25 30 35 Number of connections mobile network www.ifn.et.tu-dresden.de/tk/ 18
Simulations Results Energy, Joules 0,08 0,07 0,06 0,05 0,04 0,03 0,02 0,01 0 Energy efficiency 5 10 15 20 25 30 35 Number of connections MAC Feedback Hello mobile netwok www.ifn.et.tu-dresden.de/tk/ 19
MAC Feedback: Detection delay After 7 failed RTS retransmissions -> packet loss -> link is failed -> route is failed -> initiating route request Unpredicted detection delay www.ifn.et.tu-dresden.de/tk/ 20
MAC Feedback: Detection Delay Active neigbours: 2 Active neigbours: 5 Active neigbours: 10 Active neigbours: 15 35 cbr connections www.ifn.et.tu-dresden.de/tk/ 21
MAC: Detection Delay Complementary Cumulative Distribution Function www.ifn.et.tu-dresden.de/tk/ 22
Conclusions - I Goals Correct quick route failure detection with minimum overhead (energy, bandwidth) Route failure detection mechanisms Link level ACKs Hello messages passive ACKs (used in DSR) end-to-end acks www.ifn.et.tu-dresden.de/tk/ 23
Conclusions - II High traffic condition - detection time of MAC Feedback increases - number of MAC Feedback false decisions increases Recommendations from energy efficiency aspect for low or moderate traffic MAC Feedback for high traffic Hello messages Recommendations from goodput aspect - static network Hello messages - mobile network MAC Feedback www.ifn.et.tu-dresden.de/tk/ 24
Thank you! www.ifn.et.tu-dresden.de/tk/ 25