I’m evaluating a lightweight setup to catch efficiency drifts without buying another analytics license… Right now I’ve got Telegraf pulling eNB/gNB counters into InfluxDB with a small pandas job doing day-over-day deltas on PRB utilization and call drop rate, all running on a $300 NUC — anyone using a leaner collector or different time-series store that won’t hammer OAM bandwidth?
On a $300 NUC, swapped InfluxDB for VictoriaMetrics single-node; vmagent + 5‑min pull cut OAM about 40%. https://victoriametrics.com.
I switched the store to TimescaleDB on the same $300 NUC and let continuous aggregates handle the “day‑over‑day deltas” — killed the pandas cron and it’s been steadier. With 10‑min polling plus jitter (building on @rjohnson42’s point), OAM dropped about 30%; small caveat: bump Telegraf’s batch size or writes get chatty. If you stick with Influx, a nightly Flux task does the same; Timescale if you’re curious: https://www.timescale.com.
But > If you stick with Influx, a nightly Flux task does the same; Timescale if you’re curious: https://www.timescale.com. Agree on the nightly part — switching to 15-min PM file pulls over SFTP and letting Telegraf parse them into Influx cut OAM about 50%, and the day-over-day runs as a single Flux task; caveat: you lose sub-15-min visibility.