Transcript Trace-based Evaluation of Rate Adaptation Schemes in

Trace-based Evaluation of Rate
Adaptation Schemes in Vehicular
Environments
Kevin C. Lee
WiVeC 2010, 5/17/10
Rate Adaptation Overview
• The 802.11 a/b/g/n standards allow the use of
multiple transmission rates
– 802.11b, 4 rate options (1,2,5.5,11Mbps)
– 802.11a, 8 rate options (6,9,12,18,24,36,48,54
Mbps)
– 802.11g, 12 rate options (11a set + 11b set)
• The method to select the transmission rate in
real time is called “Rate Adaptation”
Why Rate Adaptation?
54Mbps
12Mbps
Signal is good Signal is weaker
Sender
Receiver
• Ideally, the transmission rate should be
adjusted according to the channel condition
Motivation
• MANY rate adaptation algorithms yet no fair
comparison
– Unrealistic propagation (unrealistic)
– Dynamic changing conditions (non-repeatable)
– Long system setup and device driver
implementation (impractical)
Framework & Goal
Implementation of
different application
Application Layer
Repeatable
evaluation of rate
adaptation schemes
…
Implementation of
various rate
adaptations schemes
Use field-collected
SNR to replace
synthetic value
MAC Layer
Physical Layer
Rapid deployment
independent of
hardware spec
Realistic SNR to
reflect the
environment
Physical Layer
• Collect SNR traces from moving cars
– Server broadcasts @ 6Mbps
– 2 Clients receive and record SNR
• Increase range and power of signal with an
external 7dBM antenna
• Replace SNR logic with SNR
from the field
• Derive BER and then bit error
probability
Trace Collection
• Traces from 3 different areas: City, Residential,
and Highway
Trace Collection Map
Static Traffic Route
• Car A centered at the mid point, stationary
• Car B and C move back and forth toward and
away from A
Rate Adaptation Schemes
Implementation
• RRAA-DYN adjusts rates before the current
estimation windowm
Static Traffic Route Result
• SNR from 440s
to 540s
• 40 seconds to
complete one
loop
• Signal strength
directly proportional to the distance between
them
Instantaneous Throughput for All
Algorithms
• Packet-based rate adaption schemes react
similarly to the SNR-based scheme (RAM)
• Sample rate plateau from 460-470s and 500530s
Throughput in Different Transmission
Rates
• Throughput increases with transmission rate
• ARF, RRAA-DYN, and RAM top 3
• AMRR and Sample bottom 2
Rate Distribution for All Schmes
• 6Mbps occupies the largest fraction for top 3
schemes but there are other rates => shortterm lossy channel
• Sample & AMRR can’t adapt to short-term
fluctuation
Success of ARF
• Comes from the fact that rate increases
conservatively and decreases drastically
• Not too good if the channel condition does
not change frequently
• Conclusion: Packet-based scheme does a
subpart job because of fixed parameters of
packet statistics; adaptive parameters to
improve
Impact of Environments
• Throughput degradation from residential,
highway, city; speed & traffic density play a factor
• RRAA-DYN beats
RRAA & RRAA-BASIC
=> changing transmission wind. helps
improve responsiveness
Impact of Propagation Model
• Rayleigh has higher throughput b/c it considers
fading where there is no dominant propagation
along a line of sight between transmitter and
receiver
• A more accurate
prop. model to use
b/c lead car and
trailing car are often
separated by cars in
between
Conclusion
• An integrated framework that utilizes
empirical data collected from the testbed to
objectively compare different rate adaption
schemes
– Repeatable
– Rapid
– Realistic