Help data aggregation: Mean, Median, and Percentile

Grafana InfluxDB
1

Hi, I’m using Grafana 4.6.2 with InfluxDB 1.2.4. When I created a graph for CPU’s usage_user, the differences between mean, median, and percentile are huge as you can see in my screenshot below

DnU9O2x.png1366×321 39.1 KB

My query looks like:

SELECT mean("usage_user") FROM "cpu" WHERE ("user" = 'SRV-001') AND $timeFilter GROUP BY time($__interval) fill(null)

Where data was grouped by 5 minutes.
Which one should I do to correct visualize CPU metrics? Thanks in advance

2

Hi,

Are you using telegraf for collecting cpu usage?

Marcus

3

Yes, I’m using telegraf to collect metrics.

4

I would suggest you to download community built dashboards and either use those straight away or take some inspiration to build your own.

One example which seems good for your case is this one. You can find more dashboard there if you search for telegraf and influxdb.

Marcus

5

Thanks. But the problem isn’t about dashboard, it’s the differences if I choose MEAN, MEDIAN, or PERCENTILE.

6

I don’t think percentile is any good for cpu metrics. Mean is probable what you want in this case.

Which one to choose is up to you.

Please read up on the different influxdb functions:

Marcus