320 likes | 428 Views
Multi-Rate Adaptation with Interference and Congestion Awareness. IPCCC 2011 University of California, Santa Cruz* Huawei Innovation Center^ 11/17/2011 Duy Nguyen*, J.J. Garcia-Luna-Aceves* and Cedric Westphal*^. Rate/Link Adaptation. Challenges. S. R. Limited Feedback. Interference.
E N D
Multi-Rate Adaptation with Interference and Congestion Awareness IPCCC 2011 University of California, Santa Cruz* Huawei Innovation Center^ 11/17/2011 Duy Nguyen*, J.J. Garcia-Luna-Aceves* and Cedric Westphal*^
Challenges S R Limited Feedback Interference N1 N2
Challenges Path Attenuation Multi-path Fading S R Limited Feedback Interference N1 N2
Rate Adaptation:Explicit vs Implicit Approach • Explicit: • Receiver-driven: dictates what rate that should be used • CSI S/N measurements & BER estimate • Implicit: • Sender-driven: Inferring the channel condition on the receiver • Based on RSSI measurements and ACK Packets
Rate Adaptation:Explicit vs Implicit Approach • Explicit: • Pros: measurements estimate from PHY • Cons: Incurs additional overhead, possible stale feedback due to short channel coherence time. • Implicit: • Pros: simplicity • Cons: must infer the channel condition on the receiver side
Multi-rate Adaptation with Interference Congestion Awareness (MAICA) • Implicit Approach can be very effective • Inspired by AIMD Scheme • Credit-based systems, using both packet window and time window • Allows progressive rate increase and immediate rate decrease • Reactive to changes in the environment • Compatible with current WiFi Systems
Multi-rate Adaptation with Interference Congestion Awareness (MAICA)
Multi-rate Adaptation with Interference Congestion Awareness (MAICA) Packets Bucket & Time Window Credit Bucket
Multi-rate Adaptation with Interference Congestion Awareness (MAICA) Acceptable Threshold Packets Bucket & Time Window Credit Bucket Awarding a credit
Multi-rate Adaptation with Interference Congestion Awareness (MAICA) Acceptable Threshold Packets Bucket & Time Window Credit Bucket Awarding a credit
Multi-rate Adaptation with Interference Congestion Awareness (MAICA) Acceptable Threshold Increase Rate Packets Bucket & Time Window Credit Bucket Credit has been reached
Multi-rate Adaptation with Interference Congestion Awareness (MAICA) Acceptable Threshold Decrease Rate Packets Bucket & Time Window Credit Bucket
Multi-rate Adaptation with Interference Congestion Awareness (MAICA) Acceptable Threshold Decrease Rate Packets Bucket & Time Window Credit Bucket
Multi-rate Adaptation with Interference Congestion Awareness (MAICA) More errors than success packets Decrease Rate Multiplicatively Packets Bucket & Time Window Credit Bucket
Multi-rate Adaptation with Interference Congestion Awareness (MAICA) Transmitting at a lower rate Packets Bucket & Time Window Credit Bucket
NS 3 Simulation Setup • Implicit rate adaptation evaluations • Compare against other current well-known rate adaptations such as AMRR, CARA, RRAA • Ported the popular Linux Minstrel rate adaptation to ns-3 simulations • MAICA consistently performs well in all scenarios
Scenario Setup Exponential distributed flows with mean of 3s 20m distance between each node
30 Flows and 2D Mobility in 500mx500m Topology with Random Node Placement
Fairness Evaluation Jain’s Fairness • Evaluate Jain’s Fairness Index and Aggregate Throughput • MAICA achieved fairness not at the expense of performance
Fairness with 16 Flows Static Grid Aggregate Throughput 8% gain over CARA
Fairness with 100 Flows Static Grid Aggregate Throughput 25% gain over CARA in dense networks
Conclusion • Simple and practical • Inspired by TCP and AIMD • Consistently performs well in various fading scenarios, especially in multi-user environment • Fairness achieved not at the expense of performance • Seamless integration with current Wi-Fi with Linux Kernel implementation