Opportunistic Routing and Network Lifetime Optimization

1
Opportunistic Routing and Network Lifetime
Optimization

• Rahul C. Shah
• PicoRadio Group
• BWRC Winter Retreat
• Jan 2005

2
Networking Protocol Stack
Application
Network (Routing, Addressing)
Next hop node information
Data Link (MAC, Sleep discipline, power control)
Physical
3
Exploiting Spatial Diversity Opportunistic
Routing

• Nodes know
• Their own location
• Their neighbors location

destination
Current node
4
Exploiting Spatial Diversity Opportunistic
Routing

• Nodes know
• Their own location
• Their neighbors location

Network layer specifies forwarding region
destination
MAC chooses next hop based on connectivity
Current node
5
Exploiting Spatial Diversity Opportunistic
Routing

• Nodes know
• Their own location
• Their neighbors location

Network layer specifies forwarding region
destination
MAC chooses next hop based on connectivity
Current node
6
Exploiting Spatial Diversity Opportunistic
Routing

• Nodes know
• Their own location
• Their neighbors location

Network layer specifies forwarding region
destination
MAC chooses next hop based on connectivity
Current node
7
When does opportunistic routing make sense?
(compared to geographic routing)

• For PicoRadio physical layer
• Node densities gt 9-10 neighbors/node
• Medium to bad channel quality
• Low traffic scenarios (lt1 packet/second per node)
• Improvements
• 10-20 lower power consumption
• 40 lower delay suffered by the packets
• Fewer retransmissions required ? higher goodput
  (10-15 higher for max of 3 retransmits)

R. C. Shah et. al., When does opportunistic
routing make sense?, To appear in IEEE PerSens
2005
8
Modeling Opportunistic Routing
No cross traffic ? Avg. number of packets waiting
to be forwarded in a neighborhood ? 1
9
Observations from Modeling/Analysis

• Optimal forwarding shape lens
• Use any node closer to the destination for
  forwarding
• Adjust duty cycle to have only 1 active forwarder
• For sufficiently large number of nodes (?4),
  exponential and fixed sleeping times have similar
  performance
• In general, energy spent on wakeups gt energy
  spent on forwarding packets
• All opportunistic protocols proposed till now
  have about the same power delay performance

R. C. Shah et. al., Modeling and analysis of
opportunistic routing in low traffic scenarios,
To appear in IEEE WiOpt 2005
10
Power and Delay with Wakeup Rate
Average number of neighbors/node 20 Packet
arrival rate 1.2 packets/second/neighborhood
11
Optimizing Network Lifetime

• All nodes are informed of
• Target network lifetime
• Maximum per-hop delay
• Node level control loop
• Every node individually tries to
• Achieve the target lifetime
• Maximize actual network lifetime beyond the
  target
• Network level control loop
• Network manager infrequently checks if any region
  cannot meet the target
• If a region cannot achieve target lifetime
• Add more nodes to that region
• Reduce the traffic rate in that region or relax
  the delay constraints

12
Lifetime Optimization Network Level
Lifetime constraint
Delay constraint
13
Lifetime Optimization Node Level

• Each node can observe
• Delay suffered by a packet
• Average packet rate that it serves
• Using this, it can calculate
• Total actual wakeup rate within the region
• Total packet rate within the region
• Optimal wakeup rate for the region
• The node can adjust its wakeup rate accordingly
• No communication required among the nodes

14
Duty Cycling Results
Actual total wakeup rate
Optimal wakeup rate
Number of nodes in forwarding region 4 Traffic
rate 5 pkts/sec, increased to 10 pkts/sec Target
lifetime 100 seconds
15
Conclusions

• Opportunistic routing
• Improved power and delay performance for dense
  networks, bad channels and low traffic
• Modeling and analysis has provided us with a good
  understanding of its behavior
• Detailed simulation model in Omnet in place,
  currently being implemented on TinyOS motes
• Network lifetime optimization
• Identified the network lifetime, delay and
  traffic tradeoff
• Proposed a completely decentralized mechanism to
  achieve the lifetime optimization
• Matlab simulation currently in place, currently
  being added to opportunistic routing simulation
  in Omnet