Packed time series

A packed time series stores records for multiple timepoints in each element to reduce storage space. You can create a packed time series if you have hertz data or you have numeric data that you want to compress. Both hertz and compressed numeric data must be recorded at regular intervals. Packed elements save approximately 4 bytes per record as compared to a regular time series, not including the savings for compressing the data.

For a hertz time series, each time series element is packed with records for one second. Hertz data is recorded at a regular subsecond frequency. For example, an electrical grid might have phasor measurement units that measure electrical waves at 50 hertz.

For a compressed time series, each time series element is packed with compressed records until the size of the element approaches 4 KB. All of the columns in the TimeSeries subtype must be numeric. You define compression separately for each column, except the first timestamp column, which is compressed by default. For example, a weather station might measure the wind speed, air temperature, air pressure, and precipitation for each weather sensor every 15 minutes.

Packed data is loaded faster than data that is not packed because packed data generates fewer log records. Each element that you insert into the database generates one or two log records, depending on whether logging is reduced. Packed data requires fewer elements.

When you create a virtual table or run a query, packed data is indistinguishable from time series data that is not packed. Each subelement that is shown in a virtual table has a row. Each subelement that is returned by a query is shown as an individual element.