TCP On-Off Traffic Experiment
Congestion control algorithms (CCAs) are challenged by competing on-off traffic patterns. When competitors join or leave, the traffic load of competing flows may change abruptly. CCAs should either yield bandwidth to new competitors or grab freed bandwidth resources once a competitor is left.
In the TCP on-off traffic experiment, TCP flows start and finish throughout the duration of the experiment. All TCP flows use the CCA under test. CCAs should adapt to changes in the level of competition gracefully. Maintaining a fair distribution of bandwidth among flows is a challenging task for CCAs in such a scenario.
Scenario
In the TCP on-off traffic experiment, multiple flows operate in a
static dumbbell network. Each flow generates greedy source traffic and
uses either TCP with the CCA under test or UDP. The number of flows can
be set with the parameter k. The start and the stop times
can be set for each flow individually with the parameters
start_times and stop_times, respectively.
To summarize the experiment setup:
Topology: Dumbbell topology (\(K>1\)) with static network parameters
Flows: Multiple TCP flows (\(K>1\)) that use the CCA under test with varying start and stop times
Traffic Generation Model: Greedy source traffic
Experiment Results
Experiment #94
Parameters
Command: ns3-dev-ccperf-static-dumbbell-default --experiment-name=tcp_on_off --db-path=benchmark_TcpNewReno.db '--parameters={aut:TcpNewReno,k:6,start_times:[0s,1s,2s,3s,4s,5s],stop_times:[5s,6s,7s,8s,9s,10s]}' --aut=TcpNewReno --stop-time=15s --seed=42 --start-times=0s,1s,2s,3s,4s,5s --stop-times=5s,6s,7s,8s,9s,10s --bw=96Mbps --loss=0.0 --qlen=120p --qdisc=FifoQueueDisc --rtts=15ms,15ms,15ms,15ms,15ms,15ms --sources=src_0,src_1,src_2,src_3,src_4,src_5 --destinations=dst_0,dst_1,dst_2,dst_3,dst_4,dst_5 --protocols=TCP,TCP,TCP,TCP,TCP,TCP --algs=TcpNewReno,TcpNewReno,TcpNewReno,TcpNewReno,TcpNewReno,TcpNewReno --recoveries=TcpPrrRecovery,TcpPrrRecovery,TcpPrrRecovery,TcpPrrRecovery,TcpPrrRecovery,TcpPrrRecovery '--traffic-models=Greedy(bytes=0),Greedy(bytes=0),Greedy(bytes=0),Greedy(bytes=0),Greedy(bytes=0),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 | 5.00 |
| src_1 | dst_1 | TCP | TcpNewReno | TcpPrrRecovery | Greedy(bytes=0) | 1.00 | 6.00 |
| src_2 | dst_2 | TCP | TcpNewReno | TcpPrrRecovery | Greedy(bytes=0) | 2.00 | 7.00 |
| src_3 | dst_3 | TCP | TcpNewReno | TcpPrrRecovery | Greedy(bytes=0) | 3.00 | 8.00 |
| src_4 | dst_4 | TCP | TcpNewReno | TcpPrrRecovery | Greedy(bytes=0) | 4.00 | 9.00 |
| src_5 | dst_5 | TCP | TcpNewReno | TcpPrrRecovery | Greedy(bytes=0) | 5.00 | 10.00 |
Metrics
The following tables list the flow, link, and network metrics of experiment #94. 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 | flow_2 | flow_3 | flow_4 | flow_5 | flow_6 |
|---|---|---|---|---|---|---|
| cov_in_flight_l4 | 1.41 | 1.24 | 1.05 | 1.28 | 1.08 | 1.18 |
| cov_throughput_l4 | 1.41 | 1.23 | 1.07 | 1.31 | 1.14 | 1.41 |
| flow_completion_time_l4 | 7.29 | 6.82 | 6.77 | 6.09 | 5.72 | 5.36 |
| mean_cwnd_l4 | 48.01 | 49.64 | 41.41 | 54.78 | 71.81 | 91.55 |
| mean_delivery_rate_l4 | 18.64 | 12.34 | 7.90 | 8.12 | 12.33 | 16.03 |
| mean_est_qdelay_l4 | 5.54 | 11.71 | 10.35 | 9.87 | 8.56 | 4.88 |
| mean_idt_ewma_l4 | 0.67 | 0.45 | 0.40 | 0.55 | 0.23 | 0.23 |
| mean_in_flight_l4 | 37.03 | 25.82 | 16.37 | 16.80 | 24.77 | 28.49 |
| mean_network_power_l4 | 858.71 | 529.75 | 340.72 | 350.31 | 552.30 | 828.56 |
| mean_one_way_delay_l7 | 1659.99 | 1539.75 | 1783.22 | 1431.88 | 1019.41 | 827.76 |
| mean_recovery_time_l4 | 32.97 | 32.04 | 32.70 | 36.88 | 32.18 | 35.59 |
| mean_sending_rate_l4 | 18.83 | 12.41 | 7.93 | 8.13 | 12.39 | 16.09 |
| mean_sending_rate_l7 | 18.64 | 10.12 | 7.90 | 5.48 | 12.33 | 16.03 |
| mean_srtt_l4 | 20.54 | 26.71 | 25.35 | 24.87 | 23.56 | 19.88 |
| mean_throughput_l4 | 18.64 | 12.34 | 7.90 | 8.12 | 12.33 | 16.03 |
| mean_throughput_l7 | 18.64 | 12.34 | 7.90 | 8.12 | 12.33 | 16.03 |
| mean_utility_mpdf_l4 | -0.04 | -0.05 | -0.07 | -0.12 | -0.05 | -0.05 |
| mean_utility_pf_l4 | 3.41 | 3.14 | 2.67 | 2.55 | 3.08 | 3.18 |
| mean_utilization_bdp_l4 | 0.32 | 0.22 | 0.14 | 0.15 | 0.21 | 0.25 |
| mean_utilization_bw_l4 | 0.19 | 0.13 | 0.08 | 0.08 | 0.13 | 0.17 |
| total_retransmissions_l4 | 253.00 | 91.00 | 25.00 | 17.00 | 53.00 | 47.00 |
| total_rtos_l4 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 |
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 | btl_forward |
|---|---|
| mean_qdisc_delay_l2 | 4.91 |
| mean_qdisc_length_l2 | 41.72 |
| mean_sending_rate_l1 | 65.62 |
| total_qdisc_drops_l2 | 486.00 |
Network Metrics
Network metrics assess the entire network as a
whole by aggregating other metrics, e.g., the aggregated throughput of
all flows.
Hence, the network metrics has only one column named net.
| Metric | net |
|---|---|
| mean_agg_in_flight_l4 | 149.29 |
| mean_agg_throughput_l4 | 75.37 |
| mean_agg_utility_mpdf_l4 | -0.38 |
| mean_agg_utility_pf_l4 | 18.03 |
| mean_agg_utilization_bdp_l4 | 1.29 |
| mean_agg_utilization_bw_l4 | 0.79 |
| mean_entropy_fairness_throughput_l4 | 1.38 |
| mean_jains_fairness_throughput_l4 | 0.84 |
| mean_product_fairness_throughput_l4 | 1680671.47 |
Figures
The following figures show the results of the experiment #94.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.
In Flight vs Mean Operating Point
The mean throughput and mean smoothed round-trip time (sRTT) at the transport layer of each flow. The optimal operating point is highlighted with a star (magenta). The joint operating point is given by the aggregated throughput and the mean sRTT over all flows
Distribution of the Operating Point
The empirical cumulative distribution function (eCDF) of the throughput and smoothed round-trip time (sRTT) at the transport layer of each flow.
Mean Operating Point Plane
The mean throughput and mean smoothed round-trip time (sRTT) at the transport layer of each flow.
