Design, Implementation and Evaluation of an Efficient Opportunistic Retransmission Protocol Mei-Hsuan Lu Peter Steenkiste Tsuhan Chen MobiCom 09
Outline Introduction Estimating link quality Protocol design Collision and fairness Multi-rate PRO Evaluation
Introduction PRO - Protocol for Retransmitting Opportunistically IEEE WLAN S D distance S R D
Estimating link quality Monitor success or failure of probe messages – Respond slowly to channel dynamics – Require extra bandwidth Monitor SNR of packets at receiver – – RSSI (received signal strength indicator) Noisy Th h
Estimating link quality
Protocol design Relay qualification Relay selection Relay prioritization Retransmission
Protocol design Relay qualification – Relay->destination ≠ destination->relay – Th h, on-line calibration Relay selection (eligible relay) – Broadcast “I am qualified relay!” – Select the node has highest RSSI w.r.t destination – Add node nest highest … – Until the prob. of having a node hearing source > threshold Th r
Protocol design Relay prioritization – Higher RSSI w.r.t destination -> higher priority -> smaller CW min (contention window size) Retransmission – Lack of ACK -> retransmit – Retransmission fail -> double CW, contend for channel again – Terminate: an ACK heard or retry limit reached or a new packet arrived – Re-ACK : to avoid collision, send “null” data packet
Collision and fairness Collision – Limiting number of eligible relays Fairness – More relays, more likely to gain access to channel – Mitigate unfairness: large initial CW, non uniform selection of time slot in CW
Multi-rate PRO Rate adaption – reduce packet error rate by lowering bitrates (no relay) – SampleRate : probe-based – CHARM : SNR-based Combine PRO with CHARM – Transmission failed : eligible relay retransmit when its rate ≥ source rate (having better link quality) – Aggressive rate selection
Evaluation Emulation – Static Overall Per-relay – Mobile – fairness Real world – Office building – Student lounge – g with multi-rate PRO
Emulation - static 3 environment scenarios – Freespace (outdoor) – Fading_k5 (small fading) – Fading_k0 (severe fading) 5 mechanisms – – with CHARM – with SampleRate – Mesh – Optimal PRO
Overall
Th r works well!
Per-relay
Emulation - mobile S D
Emulation - fairness S1S2D2 D1 100m S1S2D2 D1 100m50m
Real world Office building – Night Student lounge – Day – Severe fading Experiment – 10 laptops as nodes – Take turns as the source and send packet to other 9 nodes one by one – Nodes other than source and destination serve as relay
High contentionHigh fading
Real world – g with multi-rate PRO High contentionHigh fading