Title: Self Organizing Wireless Mesh Networks
1Self Organizing WirelessMesh Networks
- Microsoft Research
- March 21, 2003
Intel/Microsoft Quarterly Strategic CTO Review
2What is a Mesh Network?
e.g. MeshNetworks, Invisible Networks, Radiant
Networks, Nokias Rooftop Network
Architecture affects design decisions on Capacity
management, fairness, addressing routing,
mobility management, energy management, service
levels, integration with the Internet, etc.
3Scoping out the Problem
- What is the achievable capacity in an ideal
wireless mesh? How can we reach this optimal
capacity? - What is the best way to reach mesh nodes? That
is, how should we assign addresses and route
packets within the mesh and to the Internet? - How should we ensure fairness and privacy for
end-users and security for the network? How
should we guard against malicious nodes? - What are the applications that exploit the
properties of the Mesh?
4Mesh Formation When does a viable mesh form?
- The answer is a function of the environment, and
business model, however if we leave out the
business model
5Problem Formulation
- Question
- How many homes in the neighborhood have to sign
up before a viable mesh forms? - Answer depends on
- Definition of viable
- Neighborhood topology
- Wireless range
- Probability of participation by a given houshold
- Example Scenario
- Viable mesh group of at least 25 houses that
form a connected graph - Topology A North Seattle Neighborhood. 8214
houses, 4Km x 4Km - Wireless range 50, 100, 200 and 1000 meters
- Houses decide to join at random, independent of
each other. We consider 0.1 to 10 participation
rates.
6Mesh Formation Simulation Results
- 5-10 subscription rate needed for suburban
topologies with 200 m wireless ranges - Once a mesh forms, it is usually well-connected
- i.e. number of outliers are few (most nodes have
gt 2 neighbors) - Need to investigate other joining models
- Business model considerations will be important
Increasing range is key for viable mesh
connectivity
7Investigating current technologies
- There are many problems with existing technology
-- we cover only a few to make some points
8Background The Hidden Terminal Problem
- Consider the following scenario Tobachi75
- B is in range of A C A C are out of range
of each other - i.e. A C are hidden from each other
- A sends a packet to B
- C sends a packet to B
- The packets collide at B
- results in reduction of throughput
- CSMA doesnt work
- C cant know that it has to wait
- since it cant hear A
B
A
C
Solution RTS/CTS - with intended transmission
duration Karn90
9Multihop Networks Case Packets in Flight
Example
RTS
RTS
RTS
RTS
RTS
2
3
4
5
7
8
9
1
11
10
6
CTS
CTS
Backoff window doubles
4 nodes are active, 2 packets in flight
Backoff algorithm hurts
Microsoft Confidential
10Range and Hop Effect 802.11a 802.11b
1 wall / hop
802.11b versus 802.11a
11Conclusions from our Studies
- Multihop with IEEE 802.11a.b,g
- Severe throughput degradation as number of hops
increase - Poor fairness properties
- No guarantee that every user will get a fair
share (equal) bandwidth -
- Current software (firmware) for ad hoc 802.11
connectivity is immature - Frequent disconnects network partitioning, loss
of bcast packets
Bottom Line Current off-the-shelf WLAN
technologies are not suitable for multihop
12Overcoming Limitations, Innovating
13A 15-Node Mesh Testbed in Building 113
- IEEE 802.11a 1st generation wireless NICs
- Internally developed multihop routing protocol
- Packet overhead is minimal when nodes are
relatively static - Use it for everyday tasks, email, web, etc.
- On-going improvements in performance via
intelligent software
14Increasing Capacity Multiple Radios
Multihop wireless networks with single radio are
inefficient, as a node can not transmit and
receive simultaneously.
Network capacity can be significantly improved if
a second radio, tuned to an orthogonal channel
is available
- Multiple radios provide frequency diversity
- reduce contention
- provide robustness
15MultiRadio Unification Protocol (MUP)
- Allows systems to locally optimize use of
available spectrum - Use existing hardware
- Support legacy applications
- Interoperate with legacy hardware
- Global information should not be required
16Simulations with a Real Topology
Mesh formation among 35 randomly selected houses
252 houses in a Seattle neighborhood
Web surfer
Range is 250 meters Routes via AODV (IETF)
ITAP
17Performance using Seattle Neighborhood
Using realistic Web Traffic
40-50 reduction in delay compared to a one-radio
network
18How do Wireless Devices affect Mesh Performance
- Do we need Spectrum Etiquettes?
19In the presence of other 2.4 GHz devices
Panasonic 2.4GHz Spread Spectrum Phone 5m and 1
Wall from receiver
20Local behavior affects Global Performance!
Doesnt care
Packets get dropped!
21Summing it up
- We believe community networking will become
increasingly important. - MSR has several technologies in the works that
will make it attractive. - Viable meshes (of 25 nodes or above) can be
formed with as few as 10 of the homes
participating - Need good range and capacity - Current off-the-shelf WLAN technologies are not
suitable for building reliable high capacity
meshes - Capacity can be improved by utilizing the entire
available spectrum - Local misbehaving wireless devices cause
unacceptable performance reduction - At this time, per packet channel switching is not
a viable option. -
- Additional Notes
- Cross industry spectrum harmonization is
important for this vision to succeed. - Mesh networking is an important area of research
for MSR (researchers from Redmond, Cambridge
SVC Labs are involved).
22Backup
23Etiquette Proposal
- Transmit Power Control (TPC)
- Reduce interference between neighbors, increase
capacity through increased spatial reuse - Dynamic Frequency Selection (DFS)
- Reduce destructive interference resulting from
simultaneous transmissions - Listen Before Talk with Channel Wait Time
(LBT-CWT) - Eliminate the possibility of devices being shut
out from using the spectrum
In addition.
24Etiquette Proposal (cont.)
- TPC is applied to the entire unlicensed band
- DFS is applied to x of the unlicensed band
- LBT-CWT is applied to (100-x) of the
unlicsensed band
For example,
5 GHz Unlicensed
5.6
5.9
6.0
5.3
5.4
5.5
5.7
5.8
5.1
5.2
5.0
US
TPC, LBT-CWT
TPC, DFS
25- .... to achieve serious capacity
improvementrange, power and topology control are
necessary
Microsoft confidential
26Why Topology Control?
u
u
w
v
V
Increased Interference! Reduced throughput!
27Ensuring Connectivity while Decreasing
Interference
Who should be my neighbor ?
What should be my transmission power? Power
level influences range Power level determines
interference Power level affects routes
Want to decide locally but want to guarantee
connectivity globally
28Cone Based Algorithm
- Theorem If a ? 5p/6 and we find a neighbor in
the cone, then we are connected.
a
Transmit with minimum power within a cone till
you hit a node -- thats your power limit !
29Cone Based Algorithm with Edge Removal
Performance
Before
After