Pramana: A composable backend for networking research slashes experiment setup time

A team of 14 researchers from multiple institutions has unveiled Pramana, a composable backend specifically designed to accelerate empirical networking research. The core problem: setting up experiments to test hypotheses like 'Does a BBR download share a bottleneck fairly with Google Meet traffic?' currently requires extensive manual configuration—building traffic generators, configuring links, and collecting metrics—often starting from scratch. Pramana tackles this with an 'intent specification' that separates what data to generate (intent), where to run it (substrate), and how to produce it (mechanism). This thin-waist architecture means one specification can execute on any testbed, from cloud labs to hardware testbeds, without rewriting code.

To validate their approach, the team mined 255 distinct data-generation intents from 66 peer-reviewed networking papers. Their intent specification satisfied every single intent, whereas existing tools (e.g., Mininet, Emulab) covered at most 13%. Their current proof-of-concept already implements 34% of those intents—more than double any existing tool. The authors argue this gap will only worsen as AI-assisted ideation accelerates hypothesis generation, making a flexible data-generation backend essential. They outline a roadmap for community-driven development to close the abstraction-implementation gap, promising to reduce the lag between having an idea and generating the evidence to test it.