An Internet Heartbeat

Abstract

Obtaining sound inferences over remote networks via active or passive measurements is difficult. Active measurement campaigns face challenges of load, coverage, and visibility. Passive measurements require a privileged vantage point. Even networks under our own control too often remain poorly understood and hard to diagnose. As a step toward the democratization of Internet measurement, we consider the inferential power possible were the network to include a constant and predictable stream of dedicated lightweight measurement traffic. We posit an Internet "heartbeat," which nodes periodically send to random destinations, and show how aggregating heartbeats facilitates introspection into parts of the network that are today generally obtuse. We explore the design space of an Internet heartbeat, potential use cases, incentives, and paths to deployment.