A zero-dependency distributed columnar engine built to process, compress, and query millions of operational metrics per second with low CPU overhead.
Row storage loads full rows into memory to inspect a single column. ITInsight's Columnar Store only scans the specific column values, saving huge memory bandwidth.
SELECT SUM(latency) FROM metrics_table;Under the hood, ITInsight utilizes Run-Length Encoding (RLE) and Delta-of-Delta value offsets to store repeating metrics. Shrink your monitoring database bills by up to 90% without losing detail resolution.
Compile filters directly into SIMD CPU instructions. ITInsight scans millions of data rows in parallel using modern vector execution pipelines. Get query response times under 5ms, even on consumer hardware.
Built in systems-level Rust with standard Node.js interfaces. No external dependencies, JVM frameworks, or database servers to manage. Run it inside lightweight containers or deploy directly to edge gateways.
Ingest and compress billions of metrics from remote sensory grids.
ITInsight processes raw telemetry packet streams from industrial IoT gateways. By targeting only key time-series index bounds, it drops search times to sub-milliseconds while maintaining compact storage footprints.
// IoT Telemetry Stream Setup
const ITInsight = require("@itconverts/insight");
const client = new ITInsight.TelemetryClient({
endpoint: "https://telemetry.itconverts.com",
writeKey: process.env.IT_WRITE_KEY
});
// Stream sensor metrics every 100ms
setInterval(() => {
client.recordMetric("sensor_temperature", {
value: getSystemTemperature(),
tags: { device: "edge-grid-04b", location: "rack-12" }
});
}, 100);Real results calculated under production telemetry loads.
A comparative look at internal storage performance factors.
| Engine Feature | Row-Based Storage (MySQL/PG) | ITInsight Columnar Engine |
|---|---|---|
| Time-Series Aggregations | Requires full table scan / high resource load | ✓ Instant columnar index retrieval |
| Compression Factor | Low (1.2x - 1.5x limit due to type mixing) | ✓ High (up to 10x via vector byte-packing) |
| Sensor Event Write Latency | High (due to index rebuilding overhead) | ✓ Low (via append-only commit logs) |
| Ad-hoc Filter Performance | Slow (requires custom index creation) | ✓ Fast (native vectorized filtering) |
| Disk Footprint | Bloated (indexes exceed actual data size) | ✓ Compressed (optimized columnar pages) |
Get started with ITInsight telemetry. Lower infrastructure overhead and run query aggregations with sub-millisecond speeds.