Skip to content

Time Series

Concepts

Term Description
Daily Active Time Series The number of time series that generate new Metrics data on the day. The statistics are based on the time series that have data generated on the day. If data collection stops, no time series fees will be charged for the new day, but previously collected Metrics data can still be queried.
Measurement Represents a collection of certain statistical values, similar to a table in a relational database.
Data Point A sample of Metrics data, similar to a row in a relational database.
Time The timestamp when the data point is generated, i.e., the time when DataKit collects a certain Metrics data.
Metrics Field, stores numerical data that changes with the timestamp. For example, cpu_total, cpu_use, cpu_use_percent in the CPU Measurement.
Tags Tags, store attribute information that does not change with the timestamp. For example, host, project fields in the CPU Measurement, used to identify the actual object attributes of the Metrics.

Billing Item Statistics

The number of newly added time series is counted at hourly intervals for the day, and after obtaining 24 data points, the maximum value is taken as the actual billing quantity.

Example

Taking the CPU Measurement as an example, based on a single Metric cpu_use_pencent, there are 6 data points. Each data point contains the following fields:

  • Time field: time

  • Metric: cpu_use_pencent

  • Tags: host and project

The distribution of data points is as follows:

  • First and fourth rows: host is Hangzhou_test1, project belongs to TrueWatch, indicating the CPU usage of the Hangzhou server.

  • Second and fifth rows: host is Ningxia_test1, project belongs to TrueWatch, indicating the CPU usage of the Ningxia server.

  • Third and sixth rows: host is Singapore_test1, project belongs to TrueWatch_oversea, indicating the CPU usage of the Singapore server.

Based on the above data, there are 3 combinations of time series for the cpu_use_pencent Metric:

  1. "host":"Hangzhou_test1","project":"TrueWatch"

  2. "host":"Ningxia_test1","project":"TrueWatch"

  3. "host":"Singapore_test1","project":"TrueWatch_oversea"

To count the total number of time series for all Metrics in the current workspace, simply add the actual counted number of time series for each Metric to get the total.

Billing Formula

Daily Cost = Actual Billing Quantity / 1000 × Unit Price (apply the corresponding unit price based on the data storage strategy)

Assuming a user installs a host DataKit and enables default Metrics data collection. This host generates 600 daily active time series per day, and the cost can be estimated through the following steps:

  1. Determine the number of hosts with DataKit installed:

    For example, 1 host is installed.

  2. Calculate the number of daily active time series:

    Number of hosts × 600 = Number of daily active time series

    For example, 1 host × 600 = 600 daily active time series.

  3. Calculate the daily estimated cost:

    Unit price of the corresponding data storage strategy × Number of daily active time series / 1000

    For example, assuming the unit price is 1 yuan per thousand, the cost is 1 yuan per thousand × 600 / 1000 = 0.6 yuan.