[D] Cloud vendors define "real-time" as under a second. Control systems need under 10 milliseconds. That gap is where things break.

Been thinking about this a lot lately. Almost every analytics platform claims to support real-time processing in 2026, but most define real-time as sub-second, which is fine for dashboards and business reporting. It completely falls apart with anything that has a physical control loop.

The math is pretty unforgiving. A production line running at 400 units per minute means a unit passes the inspection point roughly every 150 milliseconds. If cloud round-trip latency is 300 to 500 milliseconds, the system flags a defect after 3 to 4 units have already passed inspection. The cloud response is technically fast by software standards. It's 30 to 50 times too slow by manufacturing standards.

The physics don't care about optimization. Data travelling through fibre moves at roughly two-thirds the speed of light. A coast-to-coast round trip in the US is around 100 milliseconds before any processing happens at all. Add routing, serialization, queuing, and TLS handshakes, and you're already past the latency budget for most control systems before computation even begins.

Curious whether teams building industrial or safety-critical systems are hitting this wall and how they're handling the architecture decision between edge, on-prem, and cloud.