Managing 100,000 network devices

The first thing that has to scale is collection. rConfig collects in parallel, queue-driven, so the work spreads across workers instead of marching device by device. Adding devices adds load the platform is designed to absorb, not a longer and longer serial run. The scale is proven, not asserted: rConfig is validated against estates of up to 200,000 simulated devices with rConfig Sim network simulator, so the throughput comes from a fleet the size of a real carrier estate.

Collecting the configs is only half the problem; you have to be able to use them. The data model holds the full, versioned history of every device across every region and stays queryable at six figures, so finding what changed, where, and when is a fast lookup rather than a batch job. Multi-tenancy keeps regions, business units, or customers separated inside one platform without standing up a separate instance for each.

Change is where scale bites hardest, because a mistake multiplies. rConfig stages and pushes validated change across the estate in controlled waves, with a known-good captured before and after, so a bad change is caught and rolled back on a few devices instead of discovered across a region. Keeping all of it backed up on a cadence that holds as the estate grows is its own discipline; see network backup for large-scale networks.