Capturing Complexity in Networked Systems Design: The Case for Improved Metrics

The systems and networking community lays great store by "clean", "elegant" system designs. Yet, our notion of what these terms mean often relies more on intuition and qualitative discussion than rigorous quantitative metric s. This paper questions whether we can do better and takes a first stab at quantifying this notion of complexity with re- gard to the algorithmic component of a networked system design. While the success of our particular attempt is unclear, we believe identifying such metrics would be valuable not only in improving our own design and analysis method- ologies but also to better articulate our design aesthetic t o other communities that design for Internet contexts ( e.g., algorithms, formal distributed systems, graph theory).