Plant-floor data, unified.
Arc is the columnar analytical database for industrial operations teams who collect telemetry from dozens of machine vendors and SCADA systems, but cannot query across them. Unify the data layer. Keep years of history. Audit-ready in SQL.
Single Go binary. Open Parquet. Standard SQL.
The shape of the problem
Every modern production facility looks the same on the inside. A line built five years ago came with one vendor's automation stack. The line built last year came with another's. The new aseptic processing system has its own historian. The energy management platform sits in its own database. The lab quality system exports to a third place.
None of them talk to each other.
When operations needs to compare OEE across both lines, someone exports CSVs. When the FSSC 22000 audit asks for a year of process data with the matching quality and traceability records, someone spends three days assembling spreadsheets. When a major customer sends a supplier scorecard, the data exists, but pulling it together is a person-week of work.
The data is not the problem. The silos are.
Arc is the unified data layer underneath. Ingest from every vendor system, every SCADA, every historian, every line. Keep years of history at full resolution. Query across all of it with one SQL surface.
Built for the workloads a real plant runs
Cross-line and cross-vendor visibility
A plant with multiple lines from multiple equipment vendors usually has multiple historians and no way to query across them. Arc ingests from all of them and stores unified telemetry as columnar Parquet. One SQL query compares yield, OEE, downtime, or process parameters across every line, regardless of which vendor collected the data.
Quality, traceability, and audit readiness
FSSC 22000, SGF IRMA, and customer audits all want the same thing: a window of process data joined to quality, traceability, and lab results. Arc keeps it all in one columnar store, queryable in SQL. Audit prep that took days becomes a query that runs in seconds.
Predictive maintenance and equipment health
Predicting failure depends on full-resolution sensor history, not downsampled averages. Vibration, temperature, current draw, valve cycles. Arc holds years of high-frequency history as ZSTD-compressed Parquet, 5-7x smaller than raw, which is what pattern models need to learn normal before they can spot what isn't.
Energy, water, and sustainability tracking
Carbon-neutral commitments need energy and water tracked per unit produced, joined to production data, kept long enough to show trends. Arc gives sustainability and operations the same data plane. Energy per liter, water per kilogram, and emissions per shift are one SQL query, not a quarterly spreadsheet.
Real-time operations dashboards
Supervisors need live dashboards on what is happening now, not yesterday's batch. Arc makes data queryable about 100 milliseconds after it lands, so Grafana, Superset, or your existing BI tool can show throughput, downtime, and quality without waiting on a nightly ETL.
What Arc gives a manufacturing operation
One data layer across every vendor
Arc speaks MQTT, OPC UA bridges, Telegraf, line protocol, REST. Whatever your existing systems can export, Arc can ingest. One place to query, instead of one place per vendor.
Audit prep in seconds, not days
When FSSC 22000, SGF IRMA, or a customer audit asks for a year of joined process, quality, and traceability data, Arc returns it as a SQL query. The alignment work happens at ingestion, not every time someone asks.
Years of history on storage you own
Open Parquet on object storage. Compaction and ZSTD compression cut raw size by 5-7x. A decade of full-resolution production data costs object-storage prices, not per-tag historian prices.
No proprietary historian lock-in
Legacy historians license per tag and store data in closed formats. Arc stores everything as open Parquet that any tool can query. If you ever stop using Arc, your archive is still readable.
Standard SQL, not a vendor DSL
Arc speaks standard SQL through DuckDB. Plant engineers, operations analysts, and consultants who know SQL are immediately productive. No proprietary tag-query language to learn.
One binary, edge to ground
One Go binary at the plant for local ingestion, the same binary in the cloud for fleet-wide visibility across sites. Same database, same SQL, same Parquet, anywhere.
Why manufacturing teams choose Arc over the alternatives
Proprietary data historians like the PI System license per tag, lock data in closed formats, and cost more every year as the plant grows. Arc is one binary, one license, open Parquet you own. Adding a new line does not add a per-tag fee.
Cloud data warehouses were built for BI, not plant-floor telemetry. Their billing model breaks when operators query in real time. Arc keeps data on storage you own and queries it without egress fees or per-query meter spikes.
Tag-indexed time-series databases hit a cardinality cliff the moment you tag readings with machine ID, line ID, batch ID, and shift. Arc treats high cardinality as ordinary column data. Fleet-wide queries are routine.
Build-it-yourself DuckDB + Parquet is the most thoughtful alternative, and it is what Arc is. DuckDB plus the production pieces a plant needs: vendor-agnostic ingestion, compaction, retention, governance, replication, and ops tooling. In one binary.
What this looks like in production
Arc is running in production manufacturing environments today, including food and beverage operations in Central America. Sensor ingestion at fleet scale. Multi-year archives in open Parquet on customer-owned object storage. SQL queries that join production, quality, and energy data across multiple lines.
AGPL-3.0 with a commercial license for organizations that need it.
Get started
Three ways to evaluate Arc for a manufacturing workload.
See it run on industrial data
Our live demos include vessel and fleet telemetry workloads, the same shape as plant-floor sensor data.
Get Arc
Arc is open source and ships as a precompiled single binary. Pull it from the download page as Docker, Helm, or a .deb, and point your existing MQTT or Telegraf pipeline at it. Most plants have data flowing in within a day. Source on GitHub.
Talk to us about your plant
For multi-line, multi-vendor environments, we run a discovery to scope ingestion, integration, and the audit and traceability use cases that matter most.