Tor Metrics Portal: Data
One of the main goals of the Tor Metrics Project is to make all gathered data available to the public. This approach enables privacy researchers to perform their own analyses using real data on the Tor network, and it acts as a safeguard to not gather data that are too sensitive to publish. The following data are available (see the Tools section for details on processing the files):
- Relay descriptor archives
- Bridge descriptor archives
- Bridge pool assignments
- Performance data
- Exit lists
The tarballs listed on this page and the raw files that were published on the last three days are also available via "rsync metrics.torproject.org::".
The relay descriptor archives contain all documents that the directory authorities make available about the network of relays. These documents include network statuses, server (relay) descriptors, and extra-info descriptors. The data formats are described here.
In order to verify the v3 votes and v3 consensuses, download the tarball of v3 certificates which is updated whenever new v3 certificates become available.
Some of the relays are configured to gather statistics on the number of requests or connecting clients, the number of processed cells per queue, or the number of exiting bytes per port. Relays running version 0.2.2.4-alpha or higher can include these statistics in extra-info descriptors, so that they are included in the relay descriptor archives. This archive contains the statistics produced by relays running earlier versions.
The bridge descriptor archives contain similar documents as the relay descriptor archives, but for the non-public bridges. The descriptors have been sanitized before publication to remove all information that could otherwise be used to locate bridges. The files below contain all documents of a given month, including bridge network statuses, bridge server descriptors, and bridge extra-info descriptors. The sanitizing process is described here.
BridgeDB periodically dumps the list of running bridges with information about the rings, subrings, and file buckets to which they are assigned to a local file. We are archiving sanitized versions of these files here to analyze how the pool assignment affects a bridge's usage.The data format and sanitizing process is described here.
We are continuously measuring the performance of the Tor network by periodically requesting files of different sizes and recording the time needed to do so. These measurements take place on moria, siv, and torperf and use an unmodified Tor client. The files below contain the output of the torperf application. The data format is described here.
The output above is the result of combining torperf request data with information about used paths. The raw files are also available below.
We further conducted additional experiments with Torperf in the past by modifying the guard node selection strategies or circuit build timeouts. The modified guard node selection strategies are to pick guard nodes from sets of the a) absolute fastest, b) absolute slowest, c) best rated vs. advertised ratio or d) worst rated vs. advertised ratio nodes. The ratio mechanisms provide a way to select the nodes that the bandwidth authorities think stand out in their measurement. Experiments are listed by the date when they ended. Details about the experiment setup are contained in a README file in the tarballs.
|April 28, 2011|
|May 18, 2011|