,
Dataset release from the WiSe deployment
|
This webpage contains information about the how to obtain the dataset collected for our MobiCom 2013 paper
Observing Home Wireless Experience through WiFi APs.
|
|
Dataset summary:
For this study, we deployed our AP based measurement infrastructure across various residential locations around Madison, Wisconsin.
We used this AP infrastructure to study the wireless properties across different types of residential wireless environments and deployments.
The deployment can be divided into three categories:-
- Building 1: This is a multi-storied apartment building within which we focused our deployment on a specific portion of the building (5 floors) having 12 units per floor.
We given away 14 APs to Building 1 residents such that each of these floors have multiple APs (usually between 2 and 4).
- Building 2: This is a 10-story building with dormitory style housing (1-2 bedrooms per apartment unit)
in which the property manager provides its own wireless service to residents by deploying its own off-the-shelf APs.
Hence, working with the property manager and the local ISP, in this building we deployed 6 APs on 2 consecutive floors (3 per floor) in the designated areas, that
replaced their existing APs.
- Others: We also distributed our APs to volunteers and colleagues staying at different low to high density residential units spread across different locations in Madison.
This was done to capture the wireless network properties across different types of home environments.
For more details regarding the measurement infrastructure and deployment, please refer to our MobiCom'13 paper.
|
|
Dataset Description
Following is the download link to the dataset (2 weeks of data from our deployment): Download dataset
Within the folder, the data collected from each AP is stored within a single CSV file (File format: <data_type>_<ap_id>.csv).
The ap_ids correspond to Building 1 (APs 1 - 14), Building 2 (APs 25 - 30) and Others (APs 15 - 24).
The data collected from each AP is organized as the following:-
Non-WiFi statistics
We use Airshark (IMC 2011 paper) to detect different non-WiFi devices operating in the vicinity of the WiSe APs.
The files are named airshark_output*.csv
| Field Name | Description |
| timestamp | The Unix timestamp when the event was recorded |
| nonwifi_dev_id | Detected non-WiFi device name |
| subband_freq | The center frequency of the 20 MHz WiFi channel in which th device was detected. |
| start_bin | Airshark reports 56 subcarrier (312.5 kHz each) energy values for each WiFi channel (excluding the 4 subcarrier values at each end of the channel). The start bin indicates the first subcarrier containing the non-WiFi device's transmissions. |
| peak_bin | The subcarrier containing the highest energy value from the non-WiFi device's transmissions. |
| end_bin | The last subcarrier in the channel containing the non-WiFi device's transmissions. |
| start_timestamp | The Unix timestamp corresponding to the start of the detected non-WiFi activity |
| end_timestamp | The Unix timestamp corresponding to the end of the detected non-WiFi activity |
| duration | The duration in seconds of the detected non-WiFi activity |
| rssi | The average RSSI of the non-WiFi device's transmissions as observed by the AP |
AP + Monitor airtime utilization statistics
AP airtime utilization statistics collected is from the first WiFi NIC inside each wireless router (which hosts the AP interface). These statistics is collected over roughly 10 second interval. For
each interval the wireless driver return the busy airtime, received airtime and transmit airtime as describe below.
The files are named ap_channel_utilization*.csv
The Monitor airtime utilization statistics table follows the same format. These statistics are collected from the second WiFi NIC inside each wireless router. This radio hops across different
WiFi channels in a round robin fashion with a dwell time of 450 - 500 milliseconds per channel. This is used to collect snapshots of the channel utilization across different WiFi channels.
The files are named phy1_channel_utilization*.csv
| Field Name | Description |
| timestamp | The Unix timestamp when the event was recorded. |
| freq | The center frequency of the WiFi channel |
| noise | Ignore, not used |
| active | The dwell time (in milliseconds) on the channel for the current stats interval |
| busy | The total time (in milliseconds) during which the energy level in the channel was above the noise floor |
| receive | The total time (in milliseconds) during which the channel was busy and a WiFi packet was detected in the channel |
| transmit | The total time (in milliseconds) during which the channel was busy and the AP itself was transmitting data |
| per_busy | busy / active |
| per_receive | receive / active |
| per_transmit | transmit / active |
Neighboring WiFi activity
We use the second radio on each Access Point to hop across different channel and collect aggregate statistics about the external WiFi activity across different channels
in a round robin fashion. The dwell time on each channel is 450 - 500 milliseconds.
The files are named external_wifi_statistics*.csv
| Field Name | Description |
| timestamp | The Unix timestamp when the event was recorded. |
| sender | The WiFi transmitter's ID |
| receiver | The WiFi receiver's ID |
| total_packets | The total number of WiFi packets transmitted by the sender in the current interval |
| total_retx | The total number of WiFi packets retransmitted by the sender in the current interval |
| avg_rate | The average PHY rate of the WiFi packets transmitted by the sender |
| channel | The WiFi channel for the link |
| num_bytes | The total number of bytes transmitted by the link. |
| retry_rate_string | Contains packet transmission statistics for each PHY rate (indicated by PHY rate * 2) concatenated by the '#' character.
For example "130@68@599#52@1@11#" indicates statistics for PHY rate 65 Mbps (130@68@599) and 26 Mbps (52@1@11). This indicates that a total of 599 packets were transmitted at 65 Mbps out which 68 packets were retried.
Similarly, 11 packets were transmitted at 26Mbps out of which 1 packet was retried. |
| link_type | Link type (uplink = 1, downlink = 2) |
AP statistics
This table provides statistics related to the AP recorded in roughly 10 second intervals.
The file are named ap_statistics*.csv.
| Field Name | Description |
| timestamp | The Unix timestamp when the event was recorded. |
noise_floor | The current noise floor setting at the AP |
| curr_freq | The current frequency used by the AP |
| crc_err | The total number of packets observed by the AP with CRC errors in the current interval |
| rx_beacons | The total number of beacons observed by the AP in the current time interval |
Client statistics
This table provides kernel statistics related to the clients associated to different WiSe APs. The statistics are recorded in roughly 10 seconds intervals
The files are named station_stats*.csv
| Field Name | Description |
| timestamp | The Unix timestamp when the event was recorded. |
| client_mac | The associated client's ID |
| last_signal | The signal strength of the client as observed at the AP |
| rate_string | Contains packet transmission statistics per PHY rate for the current 10 second time interval. Statistics for different PHY rates are concatenated by the '^' character.
For example, "1^48@332@361@38.0@97^54@15@19@36.4@84" indicates statistics for PHY rate 48 Mbps (48@332@361@38.0@97) and 54 Mbps (54@15@19@36.4@84). The PHY rate statistics for 48 Mbps indicates
that 361 packets were transmitted at 48 Mbps out of which 332 were delivered successfully. Ignore the last two fields for each rate's stats. |
| average_rate | The average PHY of all packet transmissions (including retries) |
| packet_count | The total packets transmitted in the current 10 second time interval |
| retried_packets | The total packets retried in the current interval |
Detailed packet summary traces
These traces consist of packet level summary information collected by the APs in Building 1. Using multiple APs within a single
building allowed us to collect wireless statistics from different vantage points. Following are the download links for the
traces:
Each directory contains the traces collected from different APs over the period of a day.
The files contain the data in the following format:
| Field Name | Description |
| sender | The ID of the WiFi sender |
| receiver | The ID of the WiFi receiver |
| round_number | These packet traces are divided into 15 minute rounds. The round number indicates the id of the current round.
The round numbers are synchronized in time across all WiSe APs. Thus, the data within each 15 minute round corresponds to the same 15
minute round across all WiSe APs.
|
| mac_timestamp | Microsecond level timestamp added by the wireless driver at the time when the packet reception was completed.
The is not the UNIX timestamp but a local timestamp from the wireless driver. This mac timestamp is synchronized across all WiSe APs
sniffers using techniques from PIE.
|
| phy_rate | This field contains the PHY rate * 2 used for the packet transmission. For example, 48 indicates a PHY rate of 24Mbps. |
| frame_length | The total length of the packet including header information |
| sequence_num | The sequence number of the packet |
| success | success = 1 indicated successful transmission of this MAC layer packet. 0 indicates failure |
|