CRAWDAD nyupoly/video

DASH and WebRTC video delivery over GENI WiMAX

This dataset describes measurements from Dynamic Adaptive Streaming over HTTP (DASH) and WebRTC video services, collected over the GENI WiMAX networks at NYU-Poly and UMass Amherst. These measurements are meant to elucidate the experience of an individual user of these services who is moving at walking speeds through the coverage area of a typical cellular network.

date/time of measurement start: 2013-4-20

date/time of measurement end: 2013-10-15

network configuration: These measurements were collected over the GENI WiMAX networks installed on the campuses of the Polytechnic Institute of NYU and the University of Massachusetts Amherst. Each installation includes a commercial WiMAX base station (BS) operating in a licensed frequency band, as well as other components required to route traffic from WiMAX clients to the Internet or other networks. For highly controlled experimentation, this platform has a distinct advantage over commercial cellular networks because it allows us to isolate the effects of the wireless channel quality from other variables such as competing traffic, carrier routing and shaping policies, and radio configuration.

data collection methodology: The data was collected by graduate students carrying laptop computers and following a particular predetermined path through the local campus while running the applications of interest.

Traceset

nyupoly/video/nyupoly-dash

DASH at NYU-Poly

• file: nyupoly-dash-measurements.tgz, nyupoly-dash-psnr.tgz
• description: This traceset includes measurements of network status, GPS location, and video metrics for a DASH video session conducted over the GENI WiMAX network on the campus of NYU-Poly.

• measurement purpose: Network Performance Analysis
• methodology: Measurements were collected on a Dell Inspiron Mini 1012 with an internal Intel WiMAX 6250 network adapter and a USB-connected GlobalSat BU-353 GPS receiver. 

The laptop was connected to the campus network via the WiMAX adapter, which connected to an NEC WiMAX base station located on the roof of a 4-story building at 40.6946,-73.985969. No other clients were connected to the WiMAX BS during the experiment runtime. Further details of the WiMAX setup are available in GENI WiMAX Performance: Evaluation and Comparison of Two Campus Testbeds (doi:10.1109/GREE.2013.23).

An Apache server on the campus network served 2-second DASH segments encoded at a range of bitrates from 100-6000 kbps, all with a constant resolution of 854x480p and a 24 fps frame rate. These video segments were from the "Big Buck Bunny" video in the DASH dataset (doi:10.1145/2155555.2155570).

An OML-instrumented version of the DASH-enabled VLC player v2.1.0 (http://witestlab.poly.edu/repos/omlapps/vlc/) was installed on the laptop, as was a logger application (http://witestlab.poly.edu/repos/omlapps/wimaxlogger/intel) for recording GPS and WiMAX measurements.

The measurements were collected while moving at walking speeds through the NYU-Poly campus, via the following waypoints: (40.693624,-73.985198; 40.692158,-73.985273; 40.692119,-73.984337; 40.693677,-73.98426; 40.6937,-73.984938)

nyupoly/video/umass-dash

DASH at UMass Amherst

• file: umass-dash-measurements.tgz
• description: This traceset includes measurements of network status, GPS location, and video metrics for a DASH video session conducted over the GENI WiMAX network on the campus of UMass Amherst.

• measurement purpose: Network Performance Analysis
• methodology: Measurements were collected on a Dell Latitude E5520 laptop with an internal Intel WiMAX 6250 network adapter and a USB-connected GlobalSat BU-353 GPS receiver.

The laptop was connected to the campus network via the WiMAX adapter, which connected to an NEC WiMAX base station located on the roof of a 16-story building at 42.393886, -72.527747. No other clients were connected to the WiMAX BS during the experiment runtime. Further details of the WiMAX setup are available in GENI WiMAX Performance: Evaluation and Comparison of Two Campus Testbeds (doi:10.1109/GREE.2013.23).

An Apache server on the campus network served 2-second DASH segments encoded at a range of bitrates from 100-6000 kbps, all with a constant resolution of 854x480p and a 24 fps frame rate. These video segments were from the "Big Buck Bunny" video in the DASH dataset (doi:10.1145/2155555.2155570).

An OML-instrumented version of the DASH-enabled VLC player v2.1.0 (http://witestlab.poly.edu/repos/omlapps/vlc/) was installed on the laptop, as was a logger application (http://witestlab.poly.edu/repos/omlapps/wimaxlogger/intel) for recording GPS and WiMAX measurements.
              
The measurements were collected while moving at walking speeds through the UMass Amherst campus, via the following waypoints: (42.393516,-72.530843; 42.392514,-72.531498; 42.391068,-72.531063; 42.391417,-72.52984; 42.392272,-72.530023)

nyupoly/video/umass-dash Trace

• umass-dash-measurements: Network, GPS, and video player metrics for the UMass Amherst DASH measurements.
• format: 10 sq3 files compressed in a .tgz archive.

  Each of the trace indices (1-10) represents a distinct trial. Each includes the following tables in an sqlite3 database:

* _experiment_metadata: Describes the start time of the experiment (as a Unix timestamp) and the schema of each table

* vlc_audio, vlc_video, vlc_input, and vlc_output: Contains the metrics you would see if you would open the VLC media player and select "Tools"; "Media Information"; "Statistics". The name of each field includes the data type ("i" for integer, "f" for float), the metric (e.g., "decoded audio"), and the unit (e.g., "blocks", "buffers", "bytes", "packets"). Measurements in these tables are collected at one-second intervals.

* vlc_dashDlSession: Each time a part of a video segment is downloaded by the VLC client, it reports the index of the segment (as "chunkCount"), the cumulative number of bytes and seconds spent downloading data during the entire video session ("readSession_B" and "timeSession_s" respectively), and the same metrics for the current video segment ("readChunk_B" and "timeChunk_s").

* vlc_dashRateAdaptation: Each time the VLC client prepares to download a new video segment, it records the bitrate at which it downloaded the last video segment ("empiricalRate_bps"), the buffer occupancy as a percent ("buffer_percent"; 100% represents 30 seconds of buffered video), the bitrate it uses to choose the quality level of the next segment, which is a function of both empirical rate and buffer status ("decisionRate_bps"), and the actual rate is selects to download the next segment ("chosenRate_bps")

* wimaxlogger_gps: Includes GPS measurements reported by the GPSD daemon in Linux. Measurements are collected whenever the hardware reports that a 3D fix is available. These measurements should not be considered "ground truth" since the consumer GPS device does not have great accuracy (especially in an urban environment)

* wimaxlogger_wimaxlink: Includes the network measurements reported by the Intel WiMAX driver in Linux. The units of measurement are: RSSI in dBm, CINR in dB, Frequency in kHz. Each database record also includes the following metadata: "oml_seq" is a index by table that increments by one each time a measurement is sent to the server, "oml_ts_client" describes the timestamp (as an offset from "start_time" in _experiment_metadata) at which the measurement was generated by the application, and "oml_ts_server" describes the timestamp offset at which the measurement was received by the OML server.

nyupoly/video/nyupoly-webrtc

WebRTC at NYU-Poly

• file: nyupoly-webrtc-wired-measurements.tgz, nyupoly-webrtc-wireless-measurements.tgz
• description: This traceset includes measurements of network status, GPS location, and video metrics for a WebRTC video chat conducted over the GENI WiMAX network on the campus of NYU-Poly.

• measurement purpose: Network Performance Analysis
• methodology: Measurements were collected on an HP Elitebook 2560p laptop with a USB-connected AW3 US211 WiMAX network adapter and a USB-connected GlobalSat BU-353 GPS receiver.

The laptop was connected to the campus network via the WiMAX adapter, which connected to an NEC WiMAX base station located on the roof of a 4-story building at 40.6946,-73.985969. No other clients were connected to the WiMAX BS during the experiment runtime. Further details of the WiMAX setup are available in GENI WiMAX Performance: Evaluation and Comparison of Two Campus Testbeds (doi:10.1109/GREE.2013.23).

During the experiment, we visit a WebRTC video chat webapp in Google Chrome (v27.0.1453.65). At the same time, another laptop connected to the campus network via an Ethernet link visits the same webapp and initiates a video chat session with the WiMAX-connected client. On both hosts, the WebRTC session statistics are queried at 1-second intervals using the getStats() call in the WebRTC API, and the results are sent to an OML server using the oml4js wrapper library (http://witestlab.poly.edu/repos/omlapps/oml4js/). The WiMAX-connected laptop also runs a logger application (http://witestlab.poly.edu/repos/omlapps/wimaxlogger/beceem) for recording GPS and WiMAX measurements.

The measurements were collected while the WiMAX-connected client moves at walking speeds through the NYU-Poly campus, via the following waypoints: (40.693624,-73.985198; 40.692158,-73.985273; 40.692119,-73.984337; 40.693677,-73.98426; 40.6937,-73.984938). The index of each trace (1-10) can be used to associate the two peers (wired and wireless) from the same session. However, the peers should not be considered to be time-synchronized; there is some offset between their local clocks and their timestamps should not be compared directly.

nyupoly/video/umass-webrtc

WebRTC at UMass Amherst

• file: umass-webrtc-wired-measurements.tgz, umass-webrtc-wireless-measurements.tgz
• description: This traceset includes measurements of network status, GPS location, and video metrics for a WebRTC video chat conducted over the GENI WiMAX network on the campus of UMass Amherst.

• measurement purpose: Network Performance Analysis
• methodology: Measurements were collected on a Dell Latitude E5520 laptop with an internal Intel WiMAX 6250 network adapter and a USB-connected GlobalSat BU-353 GPS receiver.

The laptop was connected to the campus network via the WiMAX adapter, which connected to an NEC WiMAX base station located on the roof of a 16-story building at 42.393886, -72.527747. No other clients were connected to the WiMAX BS during the experiment runtime. Further details of the WiMAX setup are available in GENI WiMAX Performance: Evaluation and Comparison of Two Campus Testbeds (doi:10.1109/GREE.2013.23).

During the experiment, we visit a WebRTC video chat webapp in Google Chrome (v27.0.1453.65). At the same time, another laptop connected to the campus network via an Ethernet link visits the same webapp and initiates a video chat session with the WiMAX-connected client. On both hosts, the WebRTC session statistics are queried at 1-second intervals using the getStats() call in the WebRTC API, and the results are sent to an OML server using the oml4js wrapper library (http://witestlab.poly.edu/repos/omlapps/oml4js/). The WiMAX-connected laptop also runs a logger application (http://witestlab.poly.edu/repos/omlapps/wimaxlogger/intel) for recording GPS and WiMAX measurements.

The measurements were collected while the WiMAX-connected client moves at walking speeds through the UMass Amherst campus, via the following waypoints: (42.393516,-72.530843; 42.392514,-72.531498; 42.391068,-72.531063; 42.391417,-72.52984; 42.392272,-72.530023).

The index of each trace (1-10) can be used to associate the two peers (wired and wireless) from the same session. However, the peers should not be considered to be time-synchronized; there is some offset between their local clocks and their timestamps should not be compared directly.