Large-scale monitoring of DHT traffic

Studying deployed Distributed Hash Tables (DHTs) entails monitoring DHT traffic. Commonly, DHT traffic is measured by instrumenting ordinary peers to passively record traffic. In this approach, using a small number of peers leads to a limited (and potentially biased) view of traffic. Alternatively, inserting a large number of peers may disrupt the natural traffic patterns of the DHT and lead to incorrect results. In general, accurately capturing DHT traffic is a challenging task. In this paper, we propose the idea of minimally visible monitors to capture the traffic at a large number of peers with minimum disruption to the DHT. We implement and validate our proposed technique, called Montra, on the Kad DHT. We show that Montra accurately captures around 90% of the query traffic while monitoring roughly 32,000 peers and can accurately identify destination peers for 90% of captured destination traffic. Using Montra, we characterize the traffic in Kad and present our preliminary results.