Video Conferencing Application Dataset

Nathaniel Cherian, Akhil Prasad, Sonia Fahmy, "A Microscopic View of Congestion Control Behavior in Video Conferencing Applications," In Proc. of Passive and Active Measurement Conference (PAM), March 2026.

Experimental network traffic data was collected from four Video Conferencing Applications (VCAs): Cisco Webex, Google Meet, Microsoft Teams, and Zoom, in four types of scenarios.

IP addresses were anonymized to maintain data privacy. Mapping to the organizational metadata, including company affiliation, geographic location, was done using the IPInfo API and is documented in a file named ip_company_map.csv, which also contains the aggregate global packet count.

The dataset is organized hierarchically as follows:

App
└─ NetworkScenario
└── ExperimentalTrace
└─── TraceData

The types of scenarios are:

1. ConstantBandwidth: Traces under constant link bandwidth. The format of the Experimental Trace is {app}ConstantBandwidth_{bandwidthValue(Mbps)}_{bufferSize(packets)}_{iterationNumber}.
2. ChangeBandwidth: Traces when the bandwidth is either increased or decreased. The format of the Experimental Trace is {app}ChangeBandwidth_{initialBandwidthValuel(Mbps)}-{changedBandwidthValue(Mbps)}_{bufferSize(packets)}_{iterationNumber}.
3. TCPBackground: Traces when subjected to TCP background traffic. The format of the Experimental Trace is {app}TCPBackground_{bandwidthValue(Mbps)}_{bufferSize(packets)}_{TCPCongestionControlAlgorithm}_{DownloadSize}_{iterationNumber}.
4. UDPBackground: Traces when subjected to UDP background traffic. The format of the Experimental Trace is {app}UDPBackground_{bandwidthValue(Mbps)}_{bufferSize(packets)}_{UDPFlowDuration(seconds)}_{UDPFlowBitrate(Mbps)}_{iterationNumber}.

An experimental sample comprises seven files detailing network traffic and buffer dynamics within the VCA client's BESS switch, alongside Quality of Experience (QoE) metrics.

Traffic Capture Files (tcpdump):

1. inbound_traffic_buffer_in.csv: Inbound traffic captured pre-buffer entry at the BESS switch's inbound buffer.
2. inbound_traffic_buffer_out.csv: Inbound traffic captured post-buffer exit at the BESS switch's inbound buffer.
3. outbound_traffic_buffer_in.csv: Outbound traffic captured pre-buffer entry at the BESS switch's outbound buffer.
4. outbound_traffic_buffer_out.csv: Outbound traffic captured post-buffer exit at the BESS switch's outbound buffer.

Buffer Evolution Logs:

1. inbound_traffic_buffer.log: Time-series data of the inbound buffer's state, logged upon packet entry.
2. outbound_traffic_buffer.log: Time-series data of the outbound buffer's state, logged upon packet entry.

Quality of Experience Data:

1. qoe.csv: QoE parameters measured at the receiver, derived from analysis of raw video (methodology detailed in the paper).

The data files capture network and Quality of Experience (QoE) metrics. TCP and UDP background traffic has destination IP addresses in the 10.10.5.0/24 or 10.10.6.0/24 subnets. VCA client traffic has destination IP addresses in the 10.10.7.0/24 subnet.

1. Traffic Buffer Files (*.csv):

(inbound_traffic_buffer_in.csv, inbound_traffic_buffer_out.csv, outbound_traffic_buffer_in.csv, outbound_traffic_buffer_out.csv)

Each record represents a packet within a flow and includes:

• timestamp
• 5-tuple flow identifier: src_ip, src_port, dst_ip, dst_port, protocol
• packet_length
• RTP metadata (for non-proprietary protocols): rtp payload type, rtp ssrc, rtp sequence number, rtp timestamp

2. Buffer Queue Log Files (*.log):

(inbound_traffic_buffer.log, outbound_traffic_buffer.log)

Each record logs buffer state metrics:

• Timestamp (nanoseconds)
• number of packets in the queue (Queue occupancy)
• number of packets dropped by BESS (Packet loss)

3. Quality of Experience File (qoe.csv):

Each record contains frame-level QoE metrics:

• timestamp
• qr_code_data (Frame sequence identifier)
• Timing metrics: time_sent, time_received, latency (time_received - time_sent)
• Objective video quality scores comparing original and received frames:
  • ssim_score (Structural Similarity Index Measure)
  • psnr_score (Peak Signal-to-Noise Ratio)
  • vif_score (Visual Information Fidelity)

Please click here to download the zip file (size: approximately 2.6 GB)

Please contact Akhil Prasad prasad67@purdue.edu with any questions on the paper or the dataset.

This work has been supported in part by NSF grant 2212200.