The Geometry of Tradeoffs
What Systems Design can Learn from Spacetime
For architects, decision makers, and anyone designing systems where no part sees the whole picture, but outcomes must still make sense.
How do you design systems that hold together when no part sees the whole picture? In relativity, which is our best understanding of spacetime, that question gets answered by geometry — the curvature of spacetime that provides a causal structure and coherence to what we observe. In information systems, architectures are stitched from distributed parts, each with only a partial view of data. The curvature, insofar as there is one, emerges from this tapestry. Systems cohere through tradeoffs-the geometry of a design surface shaped by constraints. A through line connects the tradeoffs-in the compromises available, in how the system behaves under their influence.
This series is an exploration of the structural logic that makes systems coherent by designing, not with omniscience, but within uncertainty, motion, and change, and where no individual component can grasp the whole picture. Drawing from physics, data systems, and hard-earned lessons in between, we will follow the deep patterns that govern distributed behavior and what they teach us about durable design.
TL; DR
This series is a technology-agnostic tour of insights gained in designing distributed systems-systems that preserve coherence under structural constraint. Coherence refers to the quality, relevance, and durability of information these systems maintain, even as parts act independently, events arrive out of order, or state changes mid-stream. Structural constraint reflects how systems, composed of interdependent components, must operate without global control or a controller.
Like in physics, where the geometry of spacetime constrains motion, what is meant by locality and causality, distributed architectures face inescapable limits shaped by space, time, and coordination cost. In this series, we explore how distributed systems remain coherent despite partial views, imperfect clocks, and costly synchronization. Whether optimizing for speed, correctness, or auditability, the goal is the same: clarity under constraint.
1. When Design Begins with Limits
I trained as an experimental particle physicist-first at Fermilab, then at CERN. Data and data systems weren’t new to me. But I was an end-user. My focus was on extracting physical insight from data, not on how the systems that managed data worked. I didn’t have to peer under the hood.
That changed after a 13-year stint and a bittersweet departure from high-octane research. My career arc brought me into IT, specifically, database operations. The role demanded more than usage. It required understanding how systems behaved internally.
What I found was strangely familiar. Not just in writing SQL, but in how data gets organized-logically and physically. The algorithms, the data structures, the constraints imposed by hardware and architecture-all bore a resemblance to patterns I had encountered in physics. I could no longer unsee them.
It’s accurate to call them databases. But the broader term is more appropriate: data systems. For this series, we’ll refer to them as systems. These include relational databases, big data platforms, pub/sub systems, document stores, streaming architectures, and other distributed models. They differ widely in purpose and application, but for this series, those distinctions matter only in part.
Continuing the story, the system had to make tradeoffs constantly: pre-compute and store results to save time later, or conserve space and compute on demand? Store more indexes and respond faster, or keep things lean and accept slower queries? There was no universal best choice. You gave something up to gain something else, whether in time or space.
As someone trained in physics, that tradeoff was all too familiar. It echoes a principle at the heart of special relativity: everything experiences motion in spacetime with a fixed overall budget. What changes is how that motion divides between space and time.
Light, for instance, moves entirely through space. From its reference frame, it experiences no passage of time. At the other extreme, an object at rest moves entirely through time, not through space. All motion is relative, but with a fixed budget. Even a stationary object, from our frame, is still “moving” through time.
That framing-space and time as a blended budget-reveals their interdependence, constrained by natural law. And that same logic appears in systems design. Use more memory to reduce latency, or conserve space and pay in time. Sometimes the tradeoffs are architectural. Sometimes they hit hard physical limits-like network latency-that are ultimate physical barriers. But the pattern endures: you don’t get to have it all. Something always gives.
It drags us-sometimes reluctantly-toward structure under constraint. Physics operates under space/time limits set by natural law. Systems do, too. But they also operate under constraints of concurrency, coordination, visibility, and cost. Whether it’s a beam of light or a distributed write in a system, coherence must emerge from partial views. A coherent light beam turns into a laser by being precise, aligned, and purposeful. A system must function coherently even when no single part sees the whole. Good design doesn’t assume omniscience. It forces us to engineer clarity in a fractured world. In that sense, tradeoffs are not bugs. They’re the geometry of what endures under those tradeoffs. Constraint isn’t just something we manage. It’s something we design within-because it shapes what’s possible.
We began by asking why systems, like spacetime, are shaped by constraints. One is imposed by geometry, the other by design, and both obey physical laws. We continue this line of inquiry over the next five installments to where it invariably leads us.
Up Next …
In the next installment, we’ll look at how the geometry of spacetime-cones, geodesics, and least action-reveals a deeper pattern. One that also governs how systems behave when no direction is cost-free, and every choice is a compromise.
Originally published at https://higgsmass.substack.com.
