Beruflich Dokumente
Kultur Dokumente
Outline
RATE ADAPTATION
General Principles of Rate Adaptation CHARM BlockRate SampleRate Robust Rate Adaptation Algorithm Rate Adaptation Algorithm Model
Definition
Exploit multi - rate capability Select best transmission rate Dynamically adapt the rate to varying channel quality No explicit feedback from receiver
Estimation
Estimate
best transmission rate from current channel conditions Infer the best transmission rate by gauging the performance of the current rate
Action
Based on estimation result , decide when and how transmission rate is updated
Deterministic
SNR
EstimationLayerSelection
EstimationMethodology Selection
Statistical
Hybrid Approach
Deterministic
Statistical
Hybrid Approach
Data Frame
Eg. Sample Rate, ARF , AARF
Signaling Frame
Eg.OAR,CA RA
HRC,Block Rate,CHA RM
Eg.
Eg.ARF , AARF
4/10/2012
PL = Ptx RSS
(2)
Destination provides its transmit power and noise level
Rate adjusted at most one level at a time Rate may jump or drop multiple levels at a time
Advantages
No use of RTS/CTS to obtain channel information and hence no RTS/CTS overhead Dependable as does not use probe packets to determine channel conditions Highly responsive to varying channel conditions as every packet used to measure channel quality
Uses historical path loss information Time-aware weighted moving average to give more weight to latest readings
Use the predicted path loss to estimate SINR at receiver Refer the SINR threshold table to determine the highest rate that satisfies the current estimated SINR
Rate Selection
Disadvantages
Per packet overhead is more Performance degrades in presence of Hidden terminals Linear weighted moving average for RSSI prediction fails in mobile environments
Observe packet success rate as a function of predicted SINR and adjust the transmission rate threshold accordingly
BlockRate
BlockRate
Rate adaptation algorithm designed for blocks Uses Multiple Rates across packets within a block that are predictive of future channel conditions BlockRate maintains a receiver-assisted technique to maintain a mapping i b between t the th SNR and d th the best b t bit bitrate t corresponding di to that SNR.
Rate
Mbps
Advantages
Amortizes overhead Responsive to dynamic channel conditions Uses different prediction models for mobile and slow-changing environments to predict SNR for packets within next block
Disadvantages
1 0 0
. . .
12 72 11
18 70 19
24 54 42
. . .
54 0 0
rcvd lost
Dependence on GPS for location information needed to predict SNR in mobile environment Cannot be used in conjunction with TCP Not suitable for delay sensitive applications like VOIP
4/10/2012
SampleRate
SampleRate
Selects the bit-rate that has the smallest average per-packet transmission time in order to achieve high throughput Periodically sends packets at bit-rates to estimate when another bit-rate will provide more throughput
Advantages
Choose higher bit-rate even if lower bit-rate has no loss in order to minimize transmission time Selects bit-rate that has smallest average per-packet transmission time in order to achieve high throughput
Reduces R d the h number b of f bit-rates b it must sample l by b eliminating l bit-rates that never send at higher rates than the current bit-rate Stops probing at a bit-rate that experiences several successive failures
Disadvantages
Calculates the average transmission time using packets sent during the last ten seconds and hence SampleRate can adapt to changing link conditions
Low responsiveness to mobile client A statistical update based on probe is too sensitive to (possible rare) failure of probe packets
Drawback : Chance events One successful probe triggers incorrect rate increase One unsuccessful probe incurs severe penalty on future rate adaptation
Guideline #4:Use PHY metrics like SNR to infer new transmission rate
Algorithms: g Charm, , HRC, , RBAR, ,
OAR
Drawbacks:
4/10/2012
Short-term loss ratio and opportunistically adapt the rate Adaptive RTS filter
Advantages
Accounts for short-term opportunistic gain Quick responsiveness Solves hidden terminal problem to some extent
MTL : a Maximum Tolerable Loss threshold PMTL P ( R ) ORI : an Opportunistic Rate Increase threshold PORI PMTL ( R ) / Ewnd: an estimation window size 1
*
Disadvantages
More overhead with RTS filter Probable solution to hidden terminal problem Needs more training and comparison to get optimized parameters
ew nd
PO R I
Block-based RSSI as a metric to characterize channel dynamics Employ Adaptive RTS to filter out collision losses with small overhead (open issue) Use short-term statistics to adapt transmission rate to dynamic channel variations Implement different prediction algorithms for varying environments
Thank You