The data displayed in this website is of two types: heatmap data and experiment data. The heatmap data is processed frequently by selecting the median throughput achieved by a given service for the last 20 experiments for a given network configuration. Therefore, the heatmap data does not represent any one specific experiment. On the other hand, the experiment data consists of all raw data from the experiment that we used to compute the heatmap data too. It can be downloaded in its zipped format and processed accordingly. Below is a further description of the parameters for the data we used in this study.

Heat Map Data

• id: Unique ID assigned by database.
• exp_id: Unique ID for an experiment that identifies two services that were pitted against each other .
• flow_id: ID to identify each service in a given experiment. For two instances of an experiment (a given exp_id) one service is assigned flow_id 1 and the other 2.
• flow_type: The category of the flow (video/web/bulk download).
• incumbent: The service whose data we are measuring .
• contender: The competing service.
• metric: The throughput achieved by the service (incumbent) in megabits per second
• exp_timestamp: The time the experiment was carried out.
• q_delay_ms: The delay in milliseconds for a queue.
• btl_bw_mbps: The experiment bandwidth setting (bottleneck bandwidth in megabits per second).
• rtt_ms: The round-trip time in milliseconds.
• queue_size_packets: The size of the packet queue.
• duration_s: The duration of the experiment in seconds.
• tar_filename: The name of a tar archive file that contains raw data for the experiment.
• testbed_version: The version of the testbed being used.
• error: The sampling error.
• fair_share: Percentage of share of median maximum-minimum fair share achieved by the service .

Experiment Data

This is a tar archive file with the following files:

• Two pcap dumps for the two services in the experiment
• One text file that starts with "Queue".

The pcap files have been used to generate the data for the heatmaps while the queue text files have been used to generate the queue graphs seen on the "Experiment History" page. The list below explains the definition of the various parameters:

• dequeued: Binary for whether a packet is inbound (0) or outbound (1).
• time: Time of experiment in nanoseconds.
• src_ip: The source IP address.
• dst_ip: The destination IP address.
• src_port: The source port number.
• dst_port: The destination port number.
• seq: The packet sequence number.
• ip_len: The length of an IP packet.
• q_size: The size of the queue.
• dropped: Binary for whether a packer is dropped(1) or not(0).
• queued: The number of packets in the queue at time t.
• batch: The number of batches.
• qid: The queue identifier.