A High-Throughput Path Metric for Multi-Hop Wireless Routing

1
A High-Throughput Path Metric for Multi-Hop
Wireless Routing

• Douglas S.J. Couto
• Daniel Aguayo
• John Bicket
• Robert Morris
• Presented by Eric Rozner

2
Overview

• Hop count metric
• Describe ETX
• Implementation
• Experiments and Comparisons
• Related Work
• Conclusions
• Discussion

3
Hop Count Metric

• Paths decided upon by shortest route
• Maximizes the distance traveled by each hop
• Minimizes signal strength -gt Maximizes the loss
  ratio
• Uses a higher TxPower -gt Interference
• Possibly many shortest routes
• Avoid lossy links?

4
Hop Count vs. Optimal
x axis throughput
y axis fraction of pairs with less throughput
5
Hop Count Route Selection
6
Motivation for a Better Routing Metric
Bidirectional loss rates
Fine-grained choices
Intermediate loss rates
7
Potential Ideas (and their cons)

• Product of per-link delivery ratios
• A perfect 2-hop route is viewed as better than a
  1-hop route with 10 loss ratio
• Throughput of a paths bottleneck link
• Same as above
• End-to-End delay
• Changes with network load as interference queue
  lengths vary can cause oscillations
• Everyone convinced?

8
ETX

• The predicted number of data transmissions
  required to send a packet over a link
• The ETX of a path is the sum of the ETX values of
  the links over that path
• Examples
• ETX of a 3-hop route with perfect links is 3
• ETX of a 1-hop route with 50 loss is 2

9
ETX continued

• Expected probability that a transmission is
  successfully received and acknowledged is df x dr
• df is forward delivery ratio
• dr is reverse delivery ratio
• Each attempt to transmit a packet is a Bernoulli
  trial, so

10
Hooray for ETX!

• Based on delivery ratios, which affect throughput
• Detects and handles asymmetry by incorporating
  loss ratios in each direction
• Uses precise link loss ratios measurements to
  make fine-grained decisions between routes
• Assumes you can measure these ratios precisely
• Penalizes routes with more hops, which have lower
  throughput due to inter-hop interference
• Assumes loss rates are generally equal over links
• Tends to minimize spectrum use, which should
  maximize overall system capacity (reduce power
  too)
• Each node spends less time retransmitting data

11
ETX always the best?
12
Acquiring ETX values

• Measured by broadcasting dedicated link probe
  packets with an average period t (jittered by
  0.1t)
• Delivery ratio
• count(t-w,t) is the of probes received during
  window w
• w/t is the of probes that should have been
  received
• Each probe contains this information

13
Implementation and such

• Authors modified DSDV and DSR
• See paper for details
• t 1 packet per second, w 10 sec
• Multiple queues (different priorities)
• Loss-ratio probes, protocol packets, data packets
• Are these experiments unfair or unrealistic?
• In DSDV w/ ETX, route table is a snapshot taken
  at end of 90 second warm-up period
• In DSR w/ ETX, source waits additional 15 sec
  before initiating the route request

14
DSDV Performance
One hop
Asymmetric
ETX inaccurate
15
DSDV and High Transmit Power
16
Packet Size Problems

• Less throughputadvantage than whendata
  packetsare smaller(134 bytes)

17
Packet sizes continued
ACKs smaller than probe packets
ETX underestimates ACK delivery ratios -gt
overestimates total number of transmissions per
packet
18
DSR Performance
Link-layer transmission feedback disabled
Link-layer transmission feedback enabled
19
Related Work

• Yarvis et al. 33 propose a path metric which
  approximates the product of the per-link delivery
  ratios.
• Link hand-shaking to avoid gray areas 8, 22
• SNR ratio is also a possible path metric 14
• SNR threshold value to filter links discovered by
  DSR Route Discovery.
• See Related Works section of Comparison of
  Routing Metrics for Static Multi-Hop Wireless
  Networks

20
Related Work Continued

• WCETT proposed in MR-LQSR
• Weighted Cumulative Expected Transmission Time

21
Conclusions

• Pros
• ETX performs better or comparable to Hop Count
  Metric
• Accounts for bi-directional loss rates
• Can easily be incorporated into routing protocols
• A lot of other work built off this paper
• Cons
• May not be best metric for all networks
• Mobility, Power-limited, Adaptive Rate
  (multi-rate), Multi-radio
• Predications of loss ratios not always accurate
• Experiments (30 sec transfer of small packets)
  may not complement real-world scenarios

22
Questions?