Loss Resilience Experiment
Random losses (e.g. packet loss due to random bit corruptions) can deteriorate the bandwidth utilization of congestion control algorithms (CCAs). Many CCAs misinterpret random losses as a signal for congestion and as a consequence the CCAs decrease the sending rate of a flow. The loss resilience experiment evaluates the influence of random losses on the operating point of CCAs.
A CCA should be resilient against random losses. There are two ways to achieve that resilience. First, a CCA may not decrease the sending rate when packet loss is detected. By doing that, a CCA can maintain its bandwidth utilization at the risk of ignoring a possible congestion signal. Second, a CCA may recover from a decrease of the sending rate quickly.
Scenario
To evaluate the loss resilience, a static dumbbell topology with a
single flow that uses greedy source traffic is set up. The experiment is
repeated with different random loss probabilities set by the parameter
loss.
To summarize the experiment setup:
Topology: Dumbbell topology (\(K=1\)) with static network parameters
Flows: A single flow (\(K=1\)) that uses a CCA
Traffic Generation Model: Greedy source traffic
Experiment Results
Experiment #7
Command: ns3-dev-ccperf-static-dumbbell-default --experiment-name=loss_resilience --db-path=benchmark_TcpNewReno.db '--parameters={aut:TcpNewReno,loss:0.02}' --aut=TcpNewReno --stop-time=15s --seed=42 --loss=0.02 --bw=16Mbps --qlen=20p --qdisc=FifoQueueDisc --rtts=15ms --sources=src_0 --destinations=dst_0 --protocols=TCP --algs=TcpNewReno --recoveries=TcpPrrRecovery --start-times=0s --stop-times=15s '--traffic-models=Greedy(bytes=0)'
Flows
| src | dst | transport_protocol | cca | cc_recovery_alg | traffic_model | start_time | stop_time |
|---|---|---|---|---|---|---|---|
| src_0 | dst_0 | TCP | TcpNewReno | TcpPrrRecovery | Greedy(bytes=0) | 0.00 | 15.00 |
Metrics
The following tables list the flow, link, and network metrics of experiment #7. Refer to the the metrics page for definitions of the listed metrics.
Flow Metrics
Flow metrics capture the performance of an individual flow. They are measured at the endpoints of a network path at either the source, the receiver, or both. Bold values indicate which flow achieved the best performance.
| Metric | flow_1 |
|---|---|
| cov_in_flight_l4 | 0.67 |
| cov_throughput_l4 | 0.88 |
| flow_completion_time_l4 | 14.93 |
| mean_cwnd_l4 | 5.90 |
| mean_delivery_rate_l4 | 3.31 |
| mean_est_qdelay_l4 | 1.05 |
| mean_idt_ewma_l4 | 5.06 |
| mean_in_flight_l4 | 5.62 |
| mean_network_power_l4 | 206.01 |
| mean_one_way_delay_l7 | 6470.69 |
| mean_recovery_time_l4 | 97.93 |
| mean_sending_rate_l4 | 3.39 |
| mean_sending_rate_l7 | 5.36 |
| mean_srtt_l4 | 16.05 |
| mean_throughput_l4 | 3.31 |
| mean_throughput_l7 | 3.22 |
| mean_utility_mpdf_l4 | -0.41 |
| mean_utility_pf_l4 | 1.17 |
| mean_utilization_bdp_l4 | 0.29 |
| mean_utilization_bw_l4 | 0.21 |
| total_retransmissions_l4 | 97.00 |
| total_rtos_l4 | 5.00 |
Link Metrics
Link metrics are recorded at the network links of interest, typically at bottlenecks. They include metrics that measure queue states. Bold values indicate which link achieved the best performance.
| Metric | link_5 |
|---|---|
| mean_qdisc_delay_l2 | 0.02 |
| mean_qdisc_length_l2 | 0.02 |
| mean_sending_rate_l1 | 3.51 |
| total_qdisc_drops_l2 | 0.00 |
Figures
The following figures show the results of the experiment #7.Time Series Plot of the Operating Point
Time series plot of the number of segments in flight, the smoothed round-trip time (sRTT), and the throughput at the transport layer.
Comparison of Congestion Control Algorithms (CCAs)
Figures
Mean Bandwidth Utilization vs Packet Error Probability
The mean bandwidth utilization vs the packet error probability. High utilization indicates good performance. Loss-resilient CCAs should reach a high bandwidth utilization independent of the packet error probability.