Every edge computing vendor wants to tell you how easy deployment is.
Zero-touch provisioning. One-click install. "Up and running in minutes." It's the headline on every landing page, the first demo in every sales call, the metric every product team optimizes for.
And honestly? They're not wrong. Deployment should be easy. But here's what most of the industry underweights: deployment is roughly 2% of the edge lifecycle.
The other 98% is runtime. And that's where the real engineering problems live.
The day after deployment
I've spent enough time in industrial environments to know that the exciting part isn't plugging in hardware and watching green lights. The exciting part is everything that comes after.
It's deciding which sensor readings get pushed upstream and which stay local. It's configuring calculation engines that correlate data across sources to detect real conditions versus noise. It's tuning filtering and aggregation rules as your operations evolve. It's keeping data intact through a three-day connectivity outage. It's making sure alarms fire and log even when the cloud is gone.
Most edge platforms have a great answer for "how do I get software onto this device?" Very few have a good answer for the engineering, data management, and operational resilience that follow — at month six, year two, day 347.
What "runtime" actually means
When I say runtime, I'm talking about everything that happens after the initial setup:
Connectivity drops — and they will. In manufacturing plants, on oil rigs, at remote substations, connectivity isn't a luxury that occasionally hiccups. It's unreliable by default. Your edge platform needs to keep working — actually working, not just "gracefully degrading" — when the cloud disappears.
Data needs to sync — without losing anything. The standard approach is store-and-forward: buffer locally, flush upstream when connectivity returns. But buffers have limits. They overflow, priorities get dropped, and nobody talks about what happens when an outage outlasts the buffer's capacity.
Alarms need to fire — even offline. If you're running OPC-UA Alarms & Conditions (and if you're in manufacturing or energy, you probably are), most platforms don't natively handle alarm lifecycle offline — acknowledgment, shelving, logging with full audit trails. That's a real gap.
Edge availability vs. data availability
The industry has made real progress on keeping edge applications running during outages. That's edge availability, and it matters.
But there's a second problem that doesn't get the same attention: data availability. Did every sensor reading make it through the outage? Is your alarm history complete? Can you prove to an auditor what happened during those three days of disconnection — not just that your app was running, but what it saw?
We started with a different question: what if connectivity never comes back? If your edge platform preserves every event in that scenario, then cloud connectivity is just upside. If it doesn't, then every outage is a potential gap in your operational record.
The compliance question
Here's where this gets concrete. An auditor walks in and says: "Show me every alarm that fired during last month's connectivity outage."
For industries governed by ISA-18.2 or similar standards, that's not a casual request. And if your edge platform wasn't capturing alarm lifecycle data — acknowledgments, shelving, state transitions — during the outage, you have a gap that no amount of application uptime can fill.
We built what we call an Event Journal to solve exactly this. It's an immutable, queryable record that captures operational data regardless of connectivity state. Not a buffer that flushes on reconnect — a persistent, audit-grade log designed for the question an auditor actually asks.
What this means for choosing an edge platform
Next time you're evaluating edge solutions, try asking these questions:
- What happens to my operational data during a multi-day connectivity outage — not my apps, my data?
- Can I query alarm history from any time period, including during outages, and get a complete record?
- How does your platform handle the difference between staying up and staying accurate?
The answers will tell you whether a platform was built for deployment day or for the years that follow.
Deployment takes a day. Runtime lasts years. We're building for the 98%.
If your edge systems have ever lost data during an outage — or you're not sure whether they have — I'd love to hear about it. We're working with a small group of POC partners in manufacturing and energy to validate our zero-data-loss architecture. Reach out if that sounds relevant.