Industrial IoT, described as engineering rather than adjectives.

What the platform connects to, why the edge-server split sits where it does, what scales cleanly, and what genuinely needs design.

Industrial IoT is a term that has survived a decade of overuse, and Indian manufacturers have reason to be sceptical of it. A large number of IIoT projects in this market produced a proof of concept, a dashboard nobody opened and an invoice. The pattern is consistent enough to be worth naming: the projects that failed were IT projects, and the ones that worked were maintenance and production projects that happened to use software.

This page describes the platform as an engineering artefact — what it connects to, how it moves and stores data, where it runs and what it secures — so it can be evaluated on architecture rather than on adjectives.

Connectivity, precisely

What the platform speaks.

LayerSupportedNotes
CNC controlsFANUC, Siemens SINUMERIK, Mitsubishi, HaasRead over the plant network using the control's own interface. Read-only.
PLCsAllen-Bradley, plus any device exposing Modbus or OPC-UATag or register level, nominated during commissioning.
Fieldbus and serialModbus RTU over RS-485, Modbus TCPCommon on energy meters, drives and older process equipment.
Discrete and analogue I/OMachineWise Gateway, optically isolated inputsFor machines with no readable control, and for added sensors.
SensorsTri-axial vibration, temperature, current, pressure, flowWired to the Gateway; sampling configured per machine class.
UplinkEthernet, Wi-Fi, 4GChosen per machine, because structured cabling rarely reaches every corner of a plant.
Edge and server, and why the split is where it is

Architecture with a reason behind it.

Acquisition happens at the edge, close to the machine, because that is where timestamps are trustworthy. A reading stamped at the machine survives a network interruption intact; a reading stamped when it arrives at a server produces a timeline that looks plausible and is wrong after every reconnection.

Aggregation, storage, computation and presentation happen on a server inside your plant. Practically, that means the system continues working during an internet outage — recording, dashboards, TV screens and local alerts all carry on. It also means your production data does not reside on infrastructure you cannot inspect, which is a recurring concern for manufacturers supplying defence, aerospace and customers with their own data clauses.

Multi-site groups keep a server per plant with a rollup that reads across them, so site-level resilience and group-level comparison are not in conflict.

Scaling, and what breaks first

Being specific about limits.

SCALES WELL

Machine count within a site

Adding machines is configuration, not a project. Dynamic nodes mean a new machine, sensor or computed data point can be added without a vendor visit or a release.

SCALES WELL

Data volume over years

Time-series storage sized for second-level data over multi-year retention, on hardware you own, with no per-gigabyte charge growing in the background.

NEEDS DESIGN

Very high-frequency sensing

Continuous raw vibration streaming from many points is a different data-volume problem from state and count monitoring. It is handled by processing spectra at the edge rather than shipping waveforms, and it needs planning rather than assumption.

NEEDS DESIGN

Sites with no reliable power

A monitoring system cannot outlive its own power supply. Where supply is unstable, the Gateway's real-time clock protects the timeline, but the server needs the same protection as any other business system.

Security and access, in practice

The risks that actually occur.

The realistic risk in manufacturing software is not an external attacker. It is the wrong people seeing commercially sensitive figures, and nobody being able to establish afterwards who saw what. Role-based access control is therefore granular by module, machine, cell and plant, with read-only viewer roles for customers or auditors, and access is logged with user attribution.

Integration with machines is read-only by design: no control parameter is modified and nothing is written to a machine. Where a plant requires a completely isolated deployment with no outbound connection, that is supported at the cost of remote support and automatic updates — a trade some defence-linked suppliers rightly choose to make.

Questions

Straight answers.

What protocols and controls does the platform support?
FANUC, Siemens SINUMERIK, Mitsubishi and Haas CNC controls read natively; Allen-Bradley and any device exposing Modbus RTU, Modbus TCP or OPC-UA; and discrete or analogue I/O through the Gateway for machines with no readable control.
Why is acquisition done at the edge?
Because timestamps taken at the machine survive a network interruption. Stamping data when it arrives at a server produces a timeline that looks plausible and is wrong after every reconnection.
Does the platform need the internet to work?
No. The server is inside your plant, so recording, dashboards, TV screens and local alerts continue through an outage. Only remote access and outbound notifications pause.
How do you handle high-frequency vibration data?
Spectra are computed at the edge rather than shipping raw waveforms to a server. Continuous high-frequency sensing across many points is a genuine data-volume design question and is planned rather than assumed.
How is adding a machine handled?
As configuration. New machines, sensors and computed data points are added on the fly without a vendor visit or a software release.
What are the real security considerations?
Mostly internal: who can see commercially sensitive figures, and whether that can be established afterwards. Access is role-based by module, machine, cell and plant, and logged with user attribution. Machine integration is read-only, and fully isolated deployment is supported where required.
FREE 2-MACHINE PILOT
Live OEE on two of your machines — this week. ₹0 to start.
Software-first setup in under an hour each · your data stays on-premises · plant-specific ROI model included.