I noticed that when using alerting with the usage-based pricing plan, cost may get quite high due to the execution of lots of check queries. Even with a few devices monitored, as soon as alerts are introduced with a low check interval of e.g. 1min, costs are rising disproportionately. I like to use a low check interval in order to get notified as soon as possible when something is wrong with a device.
Is there any better way to do this? E.g. should I use something like an edge gateway and host InfluxDB OSS there to monitor the devices?
Hello @umennel,
Is your data all in one bucket? Can you consolidate alerts with a task that unifies your data and then only run one task? What does your input data look like and what are you trying to do?
Hi @Anaisdg,
I am trying to detect anomalies in the telemetry data of a manufacturing machine. Staff has to respond ASAP if they get notified of an anomaly. That’s why I’d like to reduce the time between the occurence of an anomaly and the response from staff.
I already unified many checks into a single task. If I run that, let’s say, every minute, thats now ok.
However, if I want to go lower than that, e.g. a few seconds, costs for queries completely outweigh other costs for data transfer. If this happens, what would be the right strategy? Use InfluxDB OSS on the edge?
Thanks
Hello @umennel,
Yes that sounds like a good strategy to me. You can use OSS at the edge with small retention policies and maybe push anomaly events into cloud to obtain a broader understanding of your entire system. Additionally you might be interested in the execd telelgraf plugin. There’s a chance that you might be able to perform the anomaly detection with telegraf.
