Verdigris + Schneider Electric: Validation on EcoStruxure

Key takeaways

You're a Schneider shop. Verdigris is the validated layer that augments EcoStruxure. Schneider runs the electrical infrastructure inside more data centers than any other vendor; Verdigris adds continuous 8 kHz electrical intelligence on its own EV2 sensors across the full power path, ingests EcoStruxure telemetry alongside, and feeds intelligence back through MQTT, REST, Modbus, and BACnet emulation. Your PME dashboards keep doing what they do.

EcoStruxure and ION meters are the most widely deployed electrical-monitoring stack in data centers. Verdigris does not replace either one.
Verdigris EV2 sensors install non-invasively on circuits and capture continuous 8 kHz current and voltage waveforms across greyspace, whitespace, and the server power shelf. That is Verdigris's own physical instrumentation, vendor-neutral, spanning Schneider and non-Schneider gear on the same estate.
ION9000 meters can capture sub-cycle waveforms on demand (up to 1024 samples per cycle, ~61 kHz). That high-frequency data is event-triggered, not continuous, and aggregates to one-second polling before reaching the EcoStruxure analytics layer (PME's default Value Poll Interval — the LI_PollingPeriod_s parameter — is 30 seconds).
Verdigris becomes the system of record as the substrate. Intelligence flows back into the existing Schneider environment over MQTT, REST API, Modbus, and BACnet emulation, with zero BMS reconfiguration required.
Reliability outcome: surface hidden equipment degradation before EcoStruxure threshold alarms fire. One reference deployment with T-Mobile detected electrical degradation in ~4% of more than 800 monitored rectifiers, with a 21-day advance warning aggregated polling had missed.

What Schneider does well

Schneider Electric runs the electrical infrastructure inside more data centers than any other vendor. EcoStruxure is the connective tissue. The portfolio spans Galaxy UPS systems, ION revenue-grade meters, EcoStruxure PME, and Resource Advisor energy analytics. ION metering accuracy is a category reference. Class 0.2S devices ship into mission-critical RFPs. Hyperscale, colocation, and enterprise operators all standardize on them.

The platform's depth shows up in everyday operations. Schneider integrates across switchgear, distribution, UPS, and rack-level PDUs. EcoStruxure alarm management is mature. Facilities teams know the dashboards. They know the alarm hierarchy. They know the maintenance procedures. That accumulated familiarity is a real operational asset. Verdigris does not try to replace it. The Schneider EPMS stack is the system of record for daily operations.

Why not just deploy Schneider ION9000 at high frequency everywhere?

It's a fair question and the answer is architectural. The constraint isn't the sensor floor; it's what reaches the analytics layer where rules fire and ML models run.

ION9000 captures sub-cycle waveforms natively, at up to 1024 samples per cycle (~61 kHz at 60 Hz). That data lives at the meter for event-triggered forensic capture. It does not stream continuously into EcoStruxure. PME's default Value Poll Interval (the LI_PollingPeriod_s parameter) is 30 seconds, with Real-Time refresh also defaulting to 30 seconds. The high-frequency data is available locally for incident review, not for the analytics layer where alarms get raised. Same physical waveform, different propagation architecture.

The second part is the validated layer. Verdigris ships pretrained ML for UPS rectifier degradation, capacitor ESR drift, and harmonic-content failure signatures, trained on years of multi-site continuous waveform data. Those models don't exist on EcoStruxure. Even at full ION9000 high-frequency capture everywhere, an operator still has to build the analytics above the sensor.

You'd be paying for high-freq sensors AND still need an analytics+intelligence layer above them. Verdigris is that layer, vendor-neutral across your mixed estate.

What lives between aggregated polling samples

ION9000 meters can capture sub-cycle waveforms on demand. The hardware natively samples at up to 1024 samples per cycle (roughly 61 kHz at 60 Hz). The ION9000T variant captures transients at 10 MHz. This is real high-frequency capability at the meter.

The architectural detail that matters: that high-frequency data is event-triggered, not continuous, and it aggregates to one-second polling before reaching the EcoStruxure analytics layer. Per the Schneider PME 2023 system-performance documentation, the default Value Poll Interval (the LI_PollingPeriod_s parameter) is 30 seconds and the Real-Time refresh default is also 30 seconds. By the time a transient reaches the operator dashboard in PME or Asset Advisor, it has been collapsed into a periodic summary. One datapoint per second is the right resolution for revenue billing, demand tracking, and steady-state load balancing. It is the wrong resolution for sub-cycle phenomena once aggregation has happened upstream.

GPU training workloads ramp in roughly 8 milliseconds. UPS rectifier degradation shows up first in harmonic content. Bearing wear in cooling-system motors expresses itself through current-signature drift across 5th and 7th harmonics. That is the load-side fingerprint IEEE 519 identifies for impending failure. None of these signals survive aggregation into a one-second register at the analytics layer. Single-reading accuracy at the meter is not the issue.

This is not a Schneider limitation. It is a physics limit of the aggregated-polling stage that sits between any meter and any operator dashboard. Verdigris retains continuous 8 kHz into the analytics layer on its own EV2 sensors across the full power path: greyspace, whitespace, and the server power shelf, not only at metering points. Verdigris is also vendor-neutral: one canonical electrical model spans Schneider gear, non-Schneider gear, and mixed estates.

Schneider PME polls every 30s. Verdigris captures 240,000 samples in that window

How does Verdigris integrate with Schneider EcoStruxure?

Verdigris EV2 sensors install non-invasively on circuits, alongside the existing Schneider hardware. Clamp-on CTs capture 8 kHz waveform data continuously, per circuit, across all four DC zones: utility edge through switchgear, distribution through row and rack, and into the server power shelf. This is Verdigris's own physical instrumentation, vendor-neutral by construction.

In parallel, EcoStruxure telemetry flows into Verdigris as a reference input. ION meter data ingests over Modbus/IP. PME alarms and Asset Advisor predictive hints route in for cross-validation. Switchgear status from EcoStruxure becomes the event timeline. Each PME alarm is checked against the 8 kHz waveform feed for physics-based ranking. Data center operators receive an average of 4,484 alerts per day and ignore 67% of them due to alert fatigue; Verdigris ranks the surviving alarms by electrical risk and suppresses the noise.

Verdigris then becomes the system of record as the substrate. Intelligence flows back into the rest of the Schneider environment over four outbound surfaces:

MQTT broker (Verdigris-hosted): real-time pub/sub for validated state, alarms, and headroom limits.
REST API: programmatic access for ServiceNow, ticketing, and orchestration tools.
Modbus: production-validated OT field protocol back into SCADA and BMS.
BACnet via emulation mode: Verdigris sensors appear as native BACnet devices on the operator's network. Zero BMS reconfiguration. No integration project. The existing Schneider BMS sees Verdigris as a standard BACnet device.

Schneider keeps doing what Schneider does well. Verdigris adds the continuous high-frequency layer Schneider's architecture wasn't built to deliver into the analytics tier.

Integration surface

Direction	Schneider System	Protocol	Verdigris Use
Inbound	ION meters via PME	Modbus/IP	Reference data; cross-validated against 8 kHz waveform
Inbound	EcoStruxure PME alarms	Modbus/IP	Filtered and ranked via physics-based validation
Inbound	EcoStruxure switchgear status	Modbus/IP	Event timeline; triggers waveform deep-dive on incidents
Inbound	EcoStruxure Asset Advisor hints	API	Cross-validated against waveform-derived early warning
Outbound	Schneider BMS / SCADA	Modbus, BACnet emulation	Validated alarms and operating limits, zero BMS reconfig
Outbound	Real-time consumers	MQTT, REST API	Validated electrical state for orchestration, ticketing, dashboards

The intelligence Verdigris adds

Four capabilities live above the Schneider data plane.

Validated operating limits. Physics-based capacity headroom for every monitored circuit. The basis is continuous 8 kHz current and voltage waveforms. Verdigris holds those waveforms against power-quality envelopes derived from IEEE 519. Thermal limits come from NFPA 70. EcoStruxure polling tells you the average load across the last second. Validated operating limits tell you how close that load sits to a real failure boundary.

Early warning grounded in waveform physics. In a T-Mobile deployment, Verdigris produced a 21-day advance warning of UPS rectifier degradation. Aggregated polling had missed it. Across more than 800 monitored rectifiers, Verdigris detected electrical degradation on ~4% (about 32 units) when zero standard alarms had fired. The T-Mobile Senior Director of Infrastructure and Engineering described the operational shift:

This represents a strategic shift in how we manage critical infrastructure. While competitors react to failures, we're building proactive intelligence that prevents them.

Senior Director, Infrastructure and Engineering, T-Mobile

A second reference comes from Orange Silicon Valley, a Schneider-instrumented technology campus. Verdigris surfaced more than $40,000 in annual energy savings the existing instrumentation had not flagged. Their infrastructure engineer put it this way.

Once you start digging into the data, you find patterns that go straight to the bottom line, and this was certainly the case for us.

Scott Dworkis, Infrastructure Engineer, Orange Silicon Valley

The pattern repeats across telecom and enterprise sites. Both customers ran existing instrumentation before Verdigris.

Alert filtering. Data center operators receive an average of 4,484 alerts per day. 67% are ignored due to alert fatigue. Verdigris validates EcoStruxure alarms against the physics signature. The alarms that survive are worth a human's attention.

Single system of record across vendors. Most colocations are not pure Schneider environments. Mixed deployments are the norm. They span Eaton, Vertiv, ABB, and Siemens hardware. Plenty of sites also run Honeywell Forge or Niagara on the building side. Verdigris becomes the canonical electrical-intelligence layer across all of it. Schneider stays the operational dashboard for Schneider gear. Verdigris is the cross-vendor intelligence model that the whole estate rolls up into.

How it deploys in a Schneider shop

A colocation operator runs EcoStruxure across 50 MW of capacity. The team adds Verdigris EV2 sensors on critical circuits across switchgear, RPP, and rack-level distribution. There is no disruption to EcoStruxure. The facilities team uses the same PME dashboards, the same alarm procedures, and the same Resource Advisor reports. ION feeds ingest into Verdigris in parallel as reference data. Validated alarms and capacity headroom flow back to the existing Schneider BMS via BACnet emulation; the BMS sees Verdigris as a standard BACnet device, no reconfiguration required.

After the first 30 days, Verdigris surfaces UPS rectifiers showing harmonic degradation below the EcoStruxure alarm threshold. The team keeps the operational dashboard they already trust. They also get a validated capacity-headroom view that turns reliability risk into a finite list of circuits to inspect. Schneider stays the operational dashboard. Verdigris is the validated layer above it. The output is validated operating limits that recover capacity which would otherwise sit stranded behind conservative design margins.