Wireless Networking Using Grid

1
Wireless Networking Using Grid

• Douglas S. J. De Couto
• http//www.pdos.lcs.mit.edu/grid
• (revised)

2
Goal Build Networks from Chaos

• Constraints
• All wireless
• No centralized infrastructure
• Mobile
• Scalable

• Examples
• Rooftop networks
• Sensor networks
• Rapid deployment

3
Grid Research Problems
A to J Hello!

• Challenges
• Routing
• i. Forwarding
• ii. Path selection
• Capacity
• Power

4
Finding Good Routes

• Shortest path routing finds bad links

As max range
5
Link Quality isnt Bimodal
17 node indoor network Broadcast 4-byte UDP
packets
6
Indoor Testbed

• 17 static nodes on 5th/6th floors
• iPaq handhelds

5
5
5
6
6
5
5
5
5
6
6
5
6
5
5
5
6
wired gateway
7
Effects of Bad Links

• Problem lossy links slow throughput and reduce
  capacity
• 802.11 features positive ACKs link-level
  retransmissions
• Lost packet transmissions waste time and spectrum
• Solution choose routes other than shortest path
• A longer route may have better links
• Use different metric than hopcount

8
Proposed Route Metric Transmission Count
vs.

• Tradeoff longer route with fewer retransmits vs.
  shorter route with more retransmits
• Quantify tradeoff by estimated transmission count
  metric
• Per-packet tx count number of failed tx
  number of successful tx
• Normally exactly 1 successful tx per packet at
  each hop
• Metric features
• Route metric is sum of link metrics
• Directly measures spectrum use of route
• Estimate as tx_count 1/(fwd_rate rev_rate)

9
Geographic Forwarding
Cs radio range
A
D
F
C
G
B
E

• Packets addressed to ?idG,locationG?
• Next hop is chosen from neighbors to move packet
  geographically closer to destination location
• Per-node routing overhead constant as network
  size (nodes, area) grows
• Requires location service, which adds overhead

10
Grid Location Service (GLS)
E
H
L
B
D
J
G
A
D?
I
F
K
C

• Each node has a few servers that know its
  location
• 1. Node D sends location updates to its servers
  (B, H, K)
• 2. Node J sends a query for D to one of Ds close
  servers
• Spatial hierarchy makes GLS space and
  communications overhead O(log n)

11
The GLS Spatial Hierarchy
All nodes agree on the global origin of the grid
hierarchy All nodes agree on square sizes
12
Understanding Network Capacity

• Measure with packet-hops number of
  simultaneous packet transmissions
• Total capacity scales with number of nodes
• Spatial reuse allows capacity to scale with area
• Maximum node density ? adding nodes adds area
• Simulation result 802.11 ad hoc total capacity
  can scale
• Per-node capacity depends on communications
  patterns
• Global communication wont scale
• Local communication will (e.g. GLS)

13
Understanding Per-node Capacity

• Network provides O(n) packet-hops, n nodes
• Random communication wont scale
• Expected path length O(sqrt n) ? each packet
  uses O(sqrt n) packet-hops
• Per-node packet rate n / (n sqrt n) O(1 /
  sqrt n)
• Local communication scales
• Expected path length O(1)
• Per-node packet rate n / (n 1) O(1)
• GLS
• Expected path length O(log n)
• Per-node packet rate n / (n log n) O(1 /
  log n)

14
Grid Monitor
15
Rooftop Testbed
6
5
4
3
2
1
LCS/AI

• Omnidirectional antennas
• LCS/AI node has directional (yagi) antenna

16
Grid Summary

• Grid protocols are
• Self-configuring
• Easy to deploy
• Scalable
• Adaptable
• http//www.pdos.lcs.mit.edu/grid

17
SPAN Reducing Power Consumption

• Reduce network power consumption by turning off
  radios
• Select coordinators to stay powered on
• Maintain network connectivity
• Preserve capacity
• Routes composed of coordinator nodes
• Distributed election algorithm elects, rotates,
  and withdraws coordinators
• Simulation result network lifetime doubled