Title: A NetworkAware MAC and Routing Protocol for Effective Load Balancing in Ad Hoc Wireless Networks wit
1A Network-Aware MAC and Routing Protocol for
Effective Load Balancing in Ad Hoc Wireless
Networks with Directional AntennaSiuli Roy,
Dola Saha, Somprakash Bandyopadhyay, Tetsuro
Ueda, Shinsuke Tanaka, MobiHoc 2003
- Anand Mohanrangan
- EE 360 Presentation
- May 12th, 2004
2Background
- Routing in Ad Hoc Networks is a well studied
problem - Most routing protocols designed independent of
the physical layer (assume omnidirectional
antennas) - Considerable research into Energy saving protocols
3Directional Antennas
- Make use of directional antennas to reduce
interference and increase capacity - Mostly done at the MAC layer, not as many
directional routing protocols proposed - This paper represents some of the work done in
this field
4Issues to consider in Ad Hoc Routing
- Some of the issues include -
- Power Consumed
- Control overhead
- End-to-end delay
- Route stability (and concurrently, frequency of
route update under mobility and node failure) - Scalability
- Load Balancing
5Zone-disjoint Transmission
- Route Coupling 2 routes are close enough to
interfere with each other - Route Coupling occurs if routes S1, N1, N2, D1
and S2, N3, N4, D2 are taken - Routes are node disjoint
- Zone disjoint achieve load balancing and
reduces congestion/end-to-end delay
6System Description
- Electronically Steerable Passive Array Radiator
(ESPAR) antenna - RF Beamforming with 1 center element surrounded
by 4-6 parasitic elements in a circle - Random way-point mobility model in 2-d space
- 45 degree antenna beamwidth
7Network Awareness
- Neighborhood Link-State Table (NLSTn)
- Contains info. on neighbors and their direction
- Neighborhood Active Node List (NANLn)
- Communication activity status of neighbors
- Active Node List (ANLn)
- Perception of node n about communications
activities in the entire network - Global Link-State Table (GLSTn)
- Contains network topology information as
perceived by node n at that instant of time
8Formation of ANL/GLST
- Both contain Recency Term - Ri
- Broadcast periodically to all nodes in network
- Periodicity determines control overhead
- Accuracy vs congestion
- Fisheye approach
9MAC Protocol ( Location Tracking)
- Receiver-oriented, Rotational Sector Based
Directional MAC (using RTS/CTS with Directional
NAV) - If signal sensed above threshold ? perform 360o
sweep to obtain location information - To enable sweep, each control packet transmitted
with preceding tone (200us) - Omnidirectional RTS/CTS for benefit of
neighboring nodes ANL/GLST
10Routing Protocol Table Driven
- Used to achieve Load balancing through Maximally
Zone Disjoint Shortest Path - Use ANL to find route with lowest route
correlation factor, ? - ?ni(P) (?n?P ( ActGni a(ni ? nj) (t) ) is a
measure of the route coupling - where ACTna (t) Gna(t) n ANL(t) are the Active
Directional Neighbors of node n at transmission
zonen (a)
11Example
- Initial Route ?
- S-X-Y-Z-D
- Route not viable due to mobility in D, route
updated by Y - If next-hop not reachable by same antenna
pattern, then perform route discovery as
previously described
12Performance Evaluation - Overhead
- For TGLST 5 sec, TANL 1 sec, medium occupied
only 1.18 of time - Simulation done with 30 static nodes and static
routes
13Performance Comparisons with DSRStatic nodes
(CBR traffic, 1024 B packets)
- 5 times higher throughput
- 3.5 times less end-to-end delay
- Far fewer packet drops
- (data obtained from 20 static snap-shots)
14Performance Degradation due to MobilityLow-mobili
ty (5m/sec, 200 packets/sec)
- Higher frequency of control packets leading to
more overhead - Slight degradation due excess control traffic
15Performance of DSR with Directional Antenna (and
directional MAC) not in paper
- DSR
- Heavy Load
- Medium Load
- Light Load
- DSR w/ Directional Antenna etc
- Heavy Load
- Medium Load
- Light Load
16Conclusions - Advantages
- Performance gains obtained using directional
antennas with suitable MAC (and routing)
protocols - Energy conservation due to beamforming
- Load balancing reduces drain on nodes on
popular routes - Can reduce hop count by trading energy
conservation for greater distance
17Conclusions - Disadvantages
- Extra processing / System complexity
- In high-mobility environment, control overhead to
maintain routing tables can be very expensive - Harder to scale to 3-d terrain model
- Need to incorporate sleep times