Count samples, not just tags
One thousand tags sampled every second produce a different workload from the same tags sampled every minute. Group signals by required update interval and payload size. Include timestamps, quality fields and protocol overhead in the estimate. Keep the assumptions visible so a change in one fast signal group can be assessed.
Separate network and disk estimates
Storage compression does not necessarily reduce the bytes transmitted by the publisher. Calculate network load before applying a storage compression ratio. Add the number of stored copies and distinguish the hot query window from archival retention. Treat indexes, logs and query workspace as additional capacity requirements.
Measure representative signals
A stable temperature and a noisy vibration signal compress differently. Capture a representative period that includes transitions and abnormal operation, then measure actual storage growth using the intended settings. Record whether a deadband removes values that a future investigation would need. Lower volume is useful only if the data still serves its purpose.
Define deletion and recovery behavior
Test what happens at the retention boundary and when disk space becomes scarce. Verify backup coverage for configuration as well as measurements. Assign an owner to capacity alerts and review the growth assumption after expanding tag coverage. Keep a dated capacity worksheet beside the system design rather than relying on an informal estimate.
Put it into practice
Use the related tool to check your assumptions, then bring the results to your project discussion.