Direct Answer

Two gateways have similar clock speeds, so why can one do local alarms while disconnected and the other only pass through data? Edge-side capability rests on four things: local cache duration, programmable alarm and interlock logic, the number of local algorithms/instructions, and disconnect catch-up completeness — clock speed is only the foundation; programmable logic and cache decide whether the site "goes dark" during an outage. The ESX intelligent edge computing gateway and the CX industrial gateway, paired with the Mistudio programmable logic control software, make edge capability a configurable engineering capability.

A scene: two gateways quoted similar prices and both listed 1 GHz in the spec table — at acceptance, the network dropped for 2 hours: gateway A kept issuing local alarms and cached 15 days of data for catch-up; gateway B "went dark" — data lost, alarms silent. Same clock speed, opposite capability — edge capability was never decided by CPU clock speed; it is decided by cache, logic and algorithms.

A second scene: a factory switch room required "local trip directly when cabinet temperature exceeds 60°C" — one vendor said "1.2 GHz is fast enough", but its gateway is pure pass-through; the trip command must wait for the cloud. Another vendor's gateway runs only 800 MHz but supports local ladder-diagram programming — the trip logic lives in the gateway and executes in milliseconds. The result: the 800 MHz gateway protected the equipment; the 1.2 GHz gateway waited 2 seconds for the cloud. High clock speed does not mean capable — whether it can work locally is the truth of edge capability.

1. The Four-Dimension Edge Assessment

Local cache duration — how long data survives an outage: 15 days of cache vs 2 hours means completely different data completeness after recovery; Programmable alarm and interlock logic — whether rules can be written on the gateway (alarm above 40°C, trip on threshold), independent of the cloud; Number of local algorithms/instructions — Mistudio's 300+ instructions: compare, timer, counter, arithmetic — how complex the local logic can get; Disconnect catch-up completeness — ordered retransmission after recovery, not one record lost during the outage — the mechanism decides whether the site "goes dark".

In one sentence: clock speed decides "how fast it computes"; cache and logic decide "whether it works during an outage" — assess edge capability on all four dimensions.

The priority of the four: cache is the "bottom line" (data lost or not), logic is "capability" (can it work), algorithms are "depth" (how complex the work), and catch-up is "closure" (complete after recovery) — check in that order, secure the bottom line before discussing depth. Run the four-dimension checklist, and only order when all four check out — one missing dimension is one dimension of rework at acceptance.

A four-dimension example: a project required "7 days of data preserved through an outage and local trip on cabinet-temperature threshold even while disconnected" — checked against the four: 15-day cache (met), programmable local interlock (met), 300+ instructions (enough to write), disconnect catch-up (met) — only then was acceptance signed. Selecting by clock speed alone is leaving "does the site go dark" to luck.

Don't be fooled by the word "edge computing": some gateways call themselves "edge computing" when they are actually "edge forwarding" — data from serial port to network port, with no logic, no cache and no local judgment in between — that is not computing, it is hauling. Asking about "edge computing" is less useful than asking four numbers: cache days, instruction count, interlock channels, catch-up speed — a vendor who can answer is real edge; one who cannot is a hauler.

2. Local Cache and Disconnect Catch-Up: The Engineering Value

Cache is the "outage insurance" — when the network drops, data is stored locally first and retransmitted in order after recovery — 15 days of cache means half a month of outage loses no data; Cache priority — alarm events are kept first, routine telemetry rolls over by time window — cache space is finite, prune by value; Catch-up mechanism — after recovery, retransmit in time order; the platform deduplicates by sampling timestamp — catch-up is not "dumping garbage", it is "returning records to their places"; Engineering meaning — monthly reports and energy statistics depend on complete data — cache and catch-up decide whether the reports are "complete".

In one sentence: cache is the outage insurance and catch-up is the claim settlement — 15 days of cache plus ordered catch-up keeps reports complete through half a month of outage.

Why catch-up uses the sampling timestamp: data cached during the outage carries its sampling timestamp and returns to its place after recovery — not "dumped all at once at recovery", which would corrupt the sequence; deduplication by sampling timestamp leaves the platform with neither more nor fewer records. The engineering quality of catch-up is whether records return to the right place — catch-up aligned by sampling timestamp is what makes reports meaningful.

A cache example: a factory lost its network for 3 days; the gateway cached 15 days — after recovery, ordered catch-up left the monthly report with not one record missing; by comparison, a neighboring project's gateway cached only 2 hours — 3 days of outage lost 3 days of data, the monthly report had gaps and energy statistics distorted. The same "outage": the 15-day-cache report is complete and the 2-hour-cache report has missing corners — cache duration is the bottom line of data completeness.

Cache-storage math: 2000 data points × 1 point per second × 24 hours × 15 days ≈ 2.6 million records — gateway storage must be estimated as "data-point count × collection frequency × cache days", not "the bigger the better" but "enough by need". Plan cache by arithmetic first: points × frequency × days = storage need — then pick the cache tier; avoid overbuying waste or underbuying data loss during an outage.

3. Programmable Logic: Mistudio's 300+ Instructions and Ladder Diagrams

Ladder-diagram programming — the language electrical engineers already know, no coding background required; 300+ instructions — compare, timer, counter, arithmetic, logic combination — covering the vast majority of on-site control scenarios; Edge-side alarms — alarm rules written on the gateway fire during an outage — alarms do not depend on the cloud link; Local interlock — trip directly on cabinet-temperature threshold, millisecond response — safety actions do not wait for the cloud.

In one sentence: programmable logic turns the gateway from a "pass-through box" into an "edge brain" — Mistudio ladder diagrams plus 300+ instructions let the site's rules be written on site.

Why ladder diagrams fit the field: ladder diagrams are the electrical engineer's "mother tongue" — coils, contacts, timers, mirroring relay control logic one to one — writing edge rules in ladder diagrams means electrical staff can maintain them themselves without waiting for programmers. Mistudio lowers the programming threshold to "electrician level" — rules are changed on site, and edge logic truly lands as maintainable.

An interlock example: a switch room required "trip directly when cabinet temperature exceeds 60°C" — the rule is written in a ladder diagram on the gateway: AI reads temperature → compare 60°C → DO trips — executing in milliseconds, without going through the cloud. With a pure pass-through gateway, the trip waits for a cloud command — a 1-second network jitter means a 1-second-late trip — the safety margin is exactly that 1 second.

4. Edge Alarms vs Cloud Alarms: The Latency Division of Labor

Edge alarms — judged and triggered locally: millisecond-level, for safety actions (trips, interlocks); Cloud alarms — platform composite analysis, global judgment: second-to-minute-level, for trend-type alarms (temperature trends, harmonic degradation); Division principle — "urgent actions" go to the edge, "complex judgments" go to the cloud; Typical pairing — the edge handles "immediate danger" (threshold trip), the cloud handles "future risk" (trend warnings).

In one sentence: the edge manages "whether it is dangerous now", the cloud manages "whether it will fail later" — the latency division means each covers its own segment.

How to draw the boundary: anything that is "act when the value crosses a threshold" (trip, interlock, start/stop) goes to the edge — the faster the action, the safer; anything that "needs history, trends and global context" (degradation diagnosis, energy analysis, cross-device correlation) goes to the cloud — the more data, the more accurate. The one-line boundary: actions to the edge, analysis to the cloud — draw it wrong and either actions lag half a beat or analysis lacks data.

A division example: the same cabinet in a switch room hit 65°C — the edge's local logic tripped immediately (millisecond, protecting the equipment); meanwhile the cloud was computing the temperature trend (from 50°C to 65°C over 30 days), judged "cooling system degradation" and issued a maintenance work order. One "now" and one "future", the pair is complete protection — edge alone is firefighting, cloud alone is hindsight.

5. Edge-Capability Acceptance: The On-Site Disconnect Test

Step 1: disconnect for 30 minutes — unplug the network cable and see whether alarms still fire and data stays cached; Step 2: disconnect for 24 hours — a long outage test verifying the cache duration (e.g., 15 days); Step 3: observe recovery — plug the cable back and check whether catch-up is ordered and platform data is complete; Step 4: interlock test — simulate a threshold condition and verify millisecond-level local tripping.

In one sentence: the disconnect test is the edge capability's "health report" — 30 minutes of outage tests alarms, 24 hours tests cache, recovery tests catch-up, simulation tests interlock.

Put acceptance into the contract: write "24 hours of outage without alarm interruption or data loss, complete catch-up after recovery" into the acceptance terms — in black and white, vendors cannot be vague; verbal promises of "strong edge capability" are only proven at acceptance. Writing the disconnect-test clause into the contract turns edge capability from "marketing" into "commitment".

An acceptance example: a project accepted gateway edge capability — 2 hours disconnected: alarms still fired (local logic active), data stayed cached (cache not full); after recovery: ordered catch-up, complete platform data; simulated cabinet temperature 65°C: local trip in 0.8 seconds. All four tests passed and edge capability passed acceptance — spec tables can lie; disconnect tests cannot.

6. Back to FEXLINK: How Edge Capability Lands

To land this logic on products: the ESX intelligent edge computing gateway and the CX industrial gateway support local caching (on-demand, 15 days as standard), programmable alarm and interlock logic (Mistudio ladder diagrams with 300+ instructions), and disconnect catch-up — edge capability is a factory capability, not an option. In selection, do not ask "what is the clock speed"; ask "how long can an outage be survived, can rules be written locally, and can thresholds trip locally" — a vendor answering "15-day cache + ladder programming + millisecond interlock" treats the edge as engineering.

One summary: edge capability = cache duration (data bottom line) + programmable logic (capability ceiling) + instruction set (depth boundary) + disconnect catch-up (closure) — ask the four dimensions and clock speed is only the opening line, not the answer. Clock speed decides how fast it computes; the four dimensions decide whether it works during an outage — accept on the four, not the clock.

One reminder: edge capability is not "the stronger the better" — oversized local cache eats storage and overcomplex local logic is hard to maintain — enough is enough, sized by site need: 2 hours of cache suffices for pure monitoring, 15 days for outage-proof reports, and millisecond interlocks where safety demands. Size edge capability by "what the site needs", not by "what the spec sheet offers" — capable and reliable beats a spec-stack that is never used.

Scope and Limitations

  • The conclusions in this article are intended for LV distribution and electrical-safety sites. Criteria are compiled from public codes and FEXLINK product documentation, and product parameters follow the latest selection handbook and product knowledge base.
  • Quantitative criteria (protection zone, discharge and residual voltage, earth resistance, thresholds and grading) hold under the preconditions stated in this article. Where the site voltage level, earthing arrangement, environmental category or load profile differs, the figures must be recalculated. Specific configurations involving ESX intelligent edge computing gateway, CX industrial wristband, Mistudio programmable logic control software should be verified against actual circuit parameters.
  • Specific product models and parameters are subject to the latest selection handbook and product knowledge base; this article does not constitute selection or procurement advice.
  • Standard numbers cited here refer to the currently effective editions; when a standard is revised, the related quantitative conclusions must be re-checked.
  • Deployment and implementation for specific projects must be carried out by qualified engineers based on site survey and design drawings.

Frequently Asked Questions

Evaluating Edge Capability Beyond the Clock Speed

Two gateways have similar clock speeds, so why can one do local alarms while disconnected and the other only pass through data