Resource Allocation and Routing in Multiradio Multimode Multichannel Multirate M4 Wireless Mesh Netw

1
Resource Allocation and Routing in Multi-radio
Multi-mode Multi-channel Multi-rate (M4)Wireless
Mesh Networks

• - Optimize a multi-hop wireless community network
  by
• utilizing the capacity of all available resources
  to its full potential -
• Ting-Yu Lin

2
Talk Outline

• Background and Related Works
• Problem Statement
• Our Work M4 Wireless Mesh Network
• - Network Architecture
• - Linear Programming Model
• - Resource Allocation and Channel Assignment
  Techniques
• - Multi-channel Packet Delivery Function (mPDF)
• - Some Clarifications and Discussion
• Observations
• TINGnet Testbed
• Future Directions

3
Background and Related Works

• Features of wireless mesh networks
• - An extension of wireless multi-hop ad hoc
  networks
• - Most traffic is directed to/from Internet
  gateways
• - Static node deployment (routing is dynamic
  though)

4
(No Transcript)
5
Key Differences

• WMNs aim to diversify the capabilities of ad hoc
  networks
• Ad hoc network can be viewed as a subset of WMNs
• Several key differences
• - Wireless infrastructure/backbone
• - Integration
• - Dedicated routing and configuration
• - Multiple radios
• - Mobility

• WMNs aim to diversify the capabilities of ad hoc
  networks
• Ad hoc network can be viewed as a subset of WMNs
• Several key differences
• - Wireless infrastructure/backbone
• - Integration
• - Dedicated routing and configuration
• - Multiple radios
• - Mobility

6
Application Scenario (I)

• Eliminate dead zones
• APs replaced by mesh routers

Broadband Home Networking
7
Application Scenario (II)

• Distributed file storage
• Distributed file access
• Video streaming

Community and Neighborhood Networking
8
Application Scenario (III)

• Multiple backhaul access modems shared by all
  nodes
• Robustness and resource utilization improvement
• Expand easily as enterprise grows

Enterprise Networking
9
Critical Factors for WMNs

• Several critical factors that impact on network
  performance
• - Radio techniques
• - Scalability
• - Mesh connectivity
• - Broadband and QoS
• - Compatibility and inter-operability
• - Security
• - Ease of use

10
Background and Related Works
11
Background and Related Works
12
Problem Statement

• Whats Wrong with Minimum Hopcount?

13
Problem Statement

• Minimizing hop-count uses low-quality links
• Only a problem if many links have intermediate
  quality

14
(No Transcript)
15
(No Transcript)
16
(No Transcript)
17
Link Quality Distribution
Wide range of delivery ratios Hard to say a
link is either good or bad Forward and reverse
rates are often different
18
One Link Over 24 Hours
Packet delivery rate
Signal strength
Noise level

• Link quality varies a lot over time
• Cannot use signal-to-noise ratio to predict link
  quality

19
(No Transcript)
20
(No Transcript)
21
Motivations/Ideas Behind Our Work

• A radio connectivity map for each channel by the
  probing mechanism to periodically measure the
  wireless link capacity (bit rate loss prob.).
• Forward and reverse directions are treated as
  different links (wireless link asymmetry
  property).
• Radio channel sharing based on IEEE802.11 DCF
  contention protocol with RTS/CTS exchange
  mechanism is also modeled in our formulation
  (channel contention model).
• It would be beneficial to equip each mesh node
  with multiple radio modules, so that simultaneous
  transmissions/receiving can be enabled
  (multi-channel routing), while the network
  connectivity is appropriately preserved.
• Our ultimate goal is to maximize the mesh network
  capacity (traffic in/out of Internet gateways),
  under the restrictions of network topology
  (connectivity status), available resources
  (hardware suites, radio channels), gateway
  capabilities, and user traffic needs.

22
Our Work M4 Wireless Mesh Networking

• Our contributions
• - A global investigation on wireless
  characteristics, including available radio
  channels and effective data transmission rates
• - Resource allocations based on user traffic
  requirements and available hardware/radio
  resources
• - Enabling simultaneous traffic
  incoming/outgoing through multi-channel routing
  strategy

23
M4 Wireless Mesh Network Architecture
Internet
Internet
Internet
DSL/Cable modem
PHS
GPRS
Internet
Ethernet

• IEEE 802.11a/b/g dual-band tri-mode NICs
• Each node equipped with one or multiple NICs
• Hybrid omni-directional and directional antennas
• Heterogeneous gateways with different bandwidth
  capacities
• (cable/DSL, Ethernet, T1/E1, GPRS, PHS..)
• Static node deployment
• In reality, radio mode (11a or b or g),
  communication distance, antenna type, cable
  quality, obstacles, and channel interference all
  influence the effective transmission data rate

24
How to distribute available hardware resources,
assign radio channels, and balance traffic loads
optimally?

• Each wireless link will be associated with an
  estimated maximum bit rate over a certain channel
  in a certain mode.
• We can only afford N sets of hardware equipment
  due to budget limitation.
• Also, it is not always beneficial to equip a node
  with too many radios, for the number of
  non-interfering channels is finite.
• Assume that our network user packets are all
  aimed to access the Internet via heterogeneous
  gateways.
• Our objective is to maximize the aggregate
  throughput experienced by gateways.

25

• Given parameters
• Directed graph G (V, E)

26

• Unknown variables

27
Example Network (IEEE 802.11-based MAC)
28
Linear Programming Model
Assuming cik is known
29
Radio Channel Sharing
30
(No Transcript)
31
(No Transcript)
32
Resource Allocation and Channel Assignment
Techniques

• Decremental Interface Management (DIM)
• Incremental Interface Management (IIM)

gt To derive cik
33
Decremental Interface Management (DIM)

• Starts from the maximum number of radio
  interfaces (i.e., equip the available number of
  non-overlapping channels C NICs at each mesh
  node).
• Removes NICs step by step.
• Terminates when the total number of NICs used
  satisfies a pre-defined number N.

34
N 20, C 3
Internet
1
Currently Sni27 gt N
2
3
LP iteration 1
1
2
3
1
1
2
2
1
3
3
2
3
1
2
3
1
2
3
Removing interface involving with the least
significant traffic flow
1
1
2
2
3
3
35
N 20, C 3
Internet
1
Currently Sni26 gt N
2
3
LP iteration 2
1
2
3
1
1
2
2
1
3
3
2
3
1
2
3
1
2
3
1
1
2
2
3
36
N 20, C 3
Internet
1
Currently Sni25 gt N
2
3
LP iteration 3
1
2
3
1
1
2
2
1
3
3
2
3
1
2
3
1
2
3
1
1
2
2
37
N 20, C 3
Internet
1
Currently Sni24 gt N
2
3
LP iteration 4
1
2
3
1
2
2
1
3
3
2
3
1
2
3
1
2
3
1
1
2
2
38
N 20, C 3
Internet
1
Currently Sni23 gt N
2
3
LP iteration 5
1
2
3
1
2
2
1
3
3
2
3
1
2
3
1
3
1
1
2
2
39
N 20, C 3
Internet
1
Currently Sni22 gt N
2
3
LP iteration 6
1
2
3
1
2
2
1
3
3
3
1
2
3
1
3
1
1
2
2
40
N 20, C 3
Internet
1
Currently Sni21 gt N
2
3
LP iteration 7
1
2
3
1
2
2
1
3
3
3
1
2
3
1
1
1
2
2
41
N 20, C 3
Internet
1
Currently Sni20 N ok!
2
3
1
2
3
1
2
1
3
3
3
1
2
3
1
LP iteration 8 done!
1
1
2
2
42
Incremental Interface Management (IIM)

• Starts from one NIC equipped at each mesh node.
• Adds NICs step by step.
• Terminates when the total number of used NICs
  reaches a pre-defined number N or when LP
  optimization saturates.

43
N 20, C 3
Internet
1
Currently Sni9 lt N
LP iteration 1
1
1
1
1
1
1
1
1
44
N 20, C 3
Internet
1
Currently Sni11 lt N
LP iteration 2
1
1
1
1
1
1
1
1
45
N 20, C 3
Internet
1
Currently Sni13 lt N
LP iteration 3
1
1
1
1
1
1
1
1
46
N 20, C 3
Internet
1
Currently Sni15 lt N
LP iteration 4
1
1
1
1
1
1
1
1
47
N 20, C 3
Internet
1
Currently Sni17 lt N
LP iteration 5
1
1
1
1
1
1
1
1
48
N 20, C 3
Internet
1
Currently Sni18 lt N
LP iteration 6
1
1
1
1
1
1
1
1
49
N 20, C 3
Internet
1
Currently Sni19 lt N
LP iteration 7
1
1
1
1
1
1
1
1
50
N 20, C 3
Internet
1
Currently Sni20 N ok!
1
1
1
1
1
1
LP iteration 8 done!
1
1
51
? DIM strategy
IIM strategy ?
52
Multi-channel Packet Delivery Function (mPDF)

• Packet forwarding mechanism is re-defined to
  enable multi-channel multi-path routing.
• Data flows should be guaranteed to reach their
  destinations through heterogeneous gateways.
• Traffic dispatcher (with sufficiently large
  bandwidth) in the Network Architecture is
  required to solve the NAT problems for TCP
  connections.

53
Some Clarifications and Discussion

• If the LP formulae is unsolvable under given
  parameters, try to reduce user traffic
  requirements by the same proportion (ex 10).
  Repeat the process until the linear model is
  solvable.
• Design multiple objective functions, and solve
  the quadratic-linear programming model.
• Allow interfaces to be channel-switchable.
• How to make the MAC-layer more efficient with
  multi-channel routing? Single-channel RTS/CTS is
  NOT efficient, can we design a multi-channel MAC
  in order to better utilize channel spatial reuse
  (exposed-terminal problem can also be
  alleviated)?
• Once the number of interfaces has been decided at
  each node, we perform link status update
  periodically (ex every two hours) to
  re-calculate routing flows.

54
(No Transcript)
55
(No Transcript)
56
Testbed Deployment

• TINGnet introduction
• IEEE 802.11a/b/g network interface cards
• Currently each node equipped with one NIC
• High-sensitivity omni-directional antennas
• Heterogeneous gateways with different capacities
• (cable/DSL, campus net, T1/E1, GPRS, PHS..)
• Static node placement
• 16 nodes across about 8-10 square kilometers
  urban area

57
TINGnet (Testbed of ITRI-NCTU/NTHU Group
Meshnet)

• An IEEE 802.11-based wireless mesh network
  comprising 16 nodes, spreading across 8 km2 or so
  of Hsinchu urban area.
• Wireless broadband Internet access.
• Uses inexpensive IEEE 802.11 a/b/g radios.
• Adaptive multi-hop routing.
• Automatic configuration enabled to facilitate
  user installation.
• Like a wireless bell (?), each mesh node,
  serving as a relaying router, tings (radio
  probes) its neighbors from time to time, so that
  data packets can dynamically discover effective
  routes to reach the Internet gateways.

58
(No Transcript)
59
TINGnet Map
60
Future Directions

• Non-linear modeling for optimized solutions
• Multi-objective goal function (ex min variance
  in ?ui ?di)
• Dynamic channel assignment (switchable
  interface)
• Multi-channel MAC protocol design
• Authentication/Access control
• Extended prototype implementation
• Two-tier meshing
• Fault recovery

these are our underway research items
requires attention to hidden-terminal problem
61
Open Problems

• Whats the routing protocol?
• - path metric
• - path selection protocol
• Adaptive carrier sensing range/Tx power
  adjustment
• gt different F function and contention model
• gt network capacity further improved?

Should be distributed!
Can be centralized!
62
Comparisons with MobiCom05 paper 1 Joint
Channel Assignment and Routing for Throughput
Optimization in Multi-radio Wireless Mesh
Networks

• In 1, only uplink traffic is modeled.
• -We considered both up- and down-link traffics.
• In 1, aggregate user traffic load is given.
• -We set lower and upper bounds for each user.
• In 1, equal channel capacity is assumed.
• -We allowed different channel conditions as in
  real systems.
• In 1, gateways have unlimited bandwidth.
• -We imposed heterogeneous capacities on gateway
  nodes.
• In 1, of equipped radios is known at each
  mesh router.
• -We designed two protocols to distribute radio
  interfaces with proper channel configurations.