1. Sharp Hybrid Adaptive Routing Protocol for Mobile Ad Hoc Networks
Venugopalan Ramasubramanian (Rama)
Zygmunt Haas
Emin Gün Sirer
Cornell University

2. introduction proactive reactive
constant high overhead independent of data traffic
proactive maintenance enables low delay and often low loss rate
reactive
on-demand routing enables overhead to scale with data traffic
performance significantly affected by mobility
one protocol may out perform the other in different network conditions
change is the only constant in mobile ad hoc networks

3. sharp overview hybridization framework application specific goals
combine proactive and reactive routing
application specific goals
destination nodes independently choose adaptation goals
minimal packet overhead, target loss rate, target delay jitter
adaptability
fine grain adaptation to changing network and traffic characteristics
self-tuning and driven by analytical model
efficiency
low overhead localized mechanisms

4. sharp hybrid adaptive routing
destination
source

5. sharp hybrid adaptive routing
destination
source

6. sharp hybrid adaptive routing
destination
source

7. sharp hybrid adaptive routing
AODV
destination
source
SPR

8. sharp proactive routing (SPR)
destination rooted Directed Acyclic Graph (DAG)
multi-link routing
DAG construction
local broadcast (TTL = zone radius)
periodic reconstructions
DAG maintenance
link orientation reversal (TORA)
periodic update beacons

9. sharp proactive routing
expanding radius from r to s (r < s)
reconstruct DAG with zone radius s and TTL s
shrinking radius from r to s (r > s)
reconstruct DAG with zone radius s and TTL r
distributed coordination and reliable packet delivery not required
multiple destinations apply SPR independently
overlapping regions share overhead

10. overhead of sharp routing components
h-r h
AODV
SPR
proactive routing
independent of number of sources (S)
largely independent of mobility (λ: mean link lifetime)
depends on zone radius (r)
depends on number of nodes in proactive zone (NDr)
reactive routing
dependent on number of active routes (sources and destinations)
dependent on mobility
depends on distance (h-r)
depends on number of nodes in the search area (NSh-r)

11. sharp adaptation estimating mean link-lifetime and mean node degree
aggregated within proactive zone
piggy-backed on update beacons
estimating traffic characteristics
number of sources, routing distance, loss rate, delay jitter
measured at destination with information piggy-backed on data packets by the source
periodically choose radius before reconstruction
driven by analytical model
hysteresis
different low and high watermarks prevents oscillations

12. sharp protocols minimal packet overhead (SHARP-PO)
power and bandwidth constrained networks
estimate overhead during each reconstruction interval and adjust radius
targeted loss rate (SHARP-LR)
loss sensitive applications (TCP)
incur less overhead to achieve the target loss rate
targeted delay jitter (SHARP-DJ)
multi-media applications
incur less overhead to achieve the target delay jitter
each destination independently chooses adaptation strategy

13. evaluation GloMoSim simulator lower layers mobility scale traffic
MAC: IEEE b; range: 250m; data rate: 11Mbps
mobility
random waypoint, 0 to 20 m/s, 0 pause time
mobility fraction: fraction of mobile nodes
scale
600 nodes 3000m x 3000m, 2 pkts/sec
200 nodes 1700m x 1700m, 8 pkts/sec
traffic
single destination
multiple destinations: 1, 4, 7, 10 sources respectively
all nodes destination

14. packet overhead

15. SHARP-PO: minimal packet overhead

16. packet overhead

17. loss rate

18. SHARP-LR: target loss rate (5%)

19. SHARP-LR: packet overhead

20. loss rate

21. delay jitter

22. SHARP-DJ: target delay jitter (0.16)

23. SHARP-DJ: packet overhead

24. delay jitter

25. conclusions SHARP hybridization
explore the continuum between proactive and reactive routing strategies
application specific performance metrics
minimal packet overhead, target loss rate, target delay jitter
nodes independently choose adaptation goals
SHARP adaptability
fine grain control of hybridization across wide range of network and traffic scenarios
expensive mechanisms – clock synchronization, leader election, agreement – not required

26. SHARP vs ZRP
SHARP supports application specific adaptation strategies (loss rate, delay jitter) in addition to packet overhead
ZRP optimizes only for packet overhead
SHARP constructs proactive zones only around destination-nodes
proactive zones around all nodes in ZRP
SHARP’s proactive routing has lower overhead – only maintains routes to the center node
ZRP ‘s proactive routing is expensive – maintain multi-cast tree to the edge nodes
SHARP expands proactive zone in response to link failures, whereas ZRP shrinks the proactive zone