Glossary

Table 1 Glossary

Glossary	Description
Tags	Indicates the metadata corresponding to metrics. Tags are used to describe data sources.
Tagset	Indicates the set of Tagkey=TagValue. It can be used to define a device uniquely.
Field	Indicates the collection metrics of time-series data, such as temperature and CPU usage. It is changing with the time.
Continuous aggregation	In time series scenarios, time-series data will be performed downsampling query periodically and the query result will be stored. When users need to query the downsampling result, they can obtain it from the CA result and do not need to perform aggregation operations on original time-series data. This can improve query efficiency.
Single-value and multi-value model	Single-value model indicates that there is only one collection metric for a row of data, that is, there is only one field for a row of data. Multi-value model indicates that there are multiple collection metrics for a row of data, that is, there are multiple fields for a row of data.
Chunk	Used for managing time-series data of some devices in a chunk group. It is the actual storage unit of time-series data.
Chunk group	To efficiently delete the expired time-series data, a time-series table is usually divided into multiple chunk groups. The expired time-series data is deleted by chunk group. It is a logical unit and consists of multiple chunks.
Metrics/Fields	Indicates the metrics, such as CPU usage and idle memory.
Point/Data	Indicates the data recorded by metrics in different time points.
Downsampling	In time-series scenarios, downsampling indicates that multiple time-series data records generated in a time period are aggregated into one data record. For example, the collector generates one data record each second. After downsampling, the average value in one hour will be used.
Data imported out of order	Indicates that the time value of the data written later is less than the time value of the data written first. Out-of-order writing in time-series scenarios is usually for data with the same Tagset, that is, for the same collection device. The data reported by the same device is ordered according to time, but due to network or other reasons, the data may be out of order, that is, the time value of the data written later is greater than the time value of the data written first. Some time series databases, such as InfluxDB, have multiple entries with the same time value in the same Tagset. If there are multiple entries, there will be data overwriting, that is, the data written later will overwrite the data written first, and only the last one will be displayed during the query. In Uqbar, data with the same time value is allowed to appear in the same Tagset, and multiple data queries with the same time value will be displayed.
Time series	Indicates a metric of a device. Multiple metric values will be generated with the time. It is also called timeline.
Table	Indicates the table for storing time-series data. Users can write data into a time-series table or query data from the time-series table. A time-series table is divided into multiple chunk group by time range. Each chunk group includes multiple chunks. A time-series table can be considered as the parent table of a secondary partition table.
Retention policy	Used for defining the storage rule and retention time of time-series data.
Data imported without the time column	The time column is not specified when data is to be imported.