Reducing edge-node bandwidth usage: dynamic filtering, deadbands, and publish-on-change

Edge nodes can transmit far less data to the cloud without losing operational value by transmitting only meaningful changes rather than every sample. This article covers the four main levers — sampling vs. transmit rate, deadband/change filtering, publish-on-change, and store-and-forward prioritization — and how to apply them per tag.

Sample rate vs. transmit rate are separate settings. The node can sample a sensor at high frequency locally (for on-node analytics and buffering) while transmitting to the cloud far less often. High cloud bandwidth usage is most often caused by tags transmitting at their full local sample rate when they don't need to. Review the node's highest-frequency tags first; a single 1 Hz tag streaming continuously usually dominates the bill.

Apply a deadband (dynamic filtering). A deadband tells the node to transmit a new value only when it differs from the last transmitted value by more than a set threshold — absolute (e.g. ±0.5 bar) or percentage (e.g. ±2%). Stable analog signals such as pressures and temperatures benefit most, since a steady reading then produces almost no traffic. Set the threshold above normal sensor noise so you suppress jitter but still capture real movement.

Use publish-on-change for discrete and slow-moving tags. For digital states, statuses, and counters, configure the tag to publish only when the value changes rather than on a fixed interval. A pump running/stopped flag then sends one message per state change instead of hundreds of identical messages.

Prioritize with store-and-forward. During normal operation, route non-critical tags (trends, diagnostics) at a lower transmit rate and reserve immediate transmission for safety- and alarm-relevant tags. If the link drops, the node buffers locally and backfills on reconnect, so lowering the routine transmit rate does not lose data — it only delays non-critical history slightly.

How to apply: In the node's tag configuration, open the tag template and set the transmit interval and deadband per tag, then redeploy the template to the node. Start with the top few highest-frequency analog tags, apply a deadband, and re-check the node's data usage after a day before tuning further.

Note: Deadbands and reduced transmit rates apply to cloud transmission only. Local sampling, buffering, and on-node analytics are unaffected, so edge alarms and applications keep running at full resolution.