FlashFlow: A Secure Speed Test for Tor

FlashFlow is a new distributed bandwidth measurement system for Tor that consists of a single authority node (“coordinator”) instructing one or more measurement nodes (“measurers”) when and how to measure Tor relays. It’s results can be used directly as relay weights, or they can be used as much more accurate relay capacity estimates in another system such as TorFlow or Simple Bandwidth Scanner.

System Overview

The following diagram shows a FlashFlow deployment in action as it measurers two relays. Note how multiple measurers can work together to measure a relay (e.g. the top relay) if, for example, they do not individually possess enough bandwidth capacity to do it alone. Also note how one measurer can measure multiple relays at the same time (e.g. the bottom measurer).

_images/flashflow.svg

The coordinator instructs its Tor client to connect to the relays and tell them the measurement parameters (blue connections). The relays agree to be measured, and the coordinator tells the appropriate measurers (over the green connections) to connect to the relays with many TLS connections. The measurers instruct their Tor clients to do so (thus establishing red connections), and when done, inform the coordinator of their success over the green connections. Once all red connections are established for the measurement of a relay, the coordinator tells the measurers to begin measuring, they forward this command to their Tor clients, and measurement traffic starts getting sent to the relay.

Relays echo measurement traffic back to measurers on the red connections, and for every second during the measurement, report to the coordinator on the blue connections how much background client traffic they forwarded in the preceding second. When a relay sends its last background traffic report, it closes all red connections and the blue connection.

The coordinator logs all per-second measurement data from measurers and relays into an intermediate results file. Periodically it generates a v3bw file using the latest measurement results for each relay. More specifically, the coordinator sums the per-second reported throughputs into 30 sums (assuming a 30 second measurement) and uses the median of these sums as the weight for each relay listed in the v3bw file.

More information on these processes can be found here:

Indices and tables