Buffering

The reason ARDIs version of graphs are known as funnels is that they can buffer data.

This allows them to store more than just the most recent values - they can keep multiple (even hours) of data that may be needed to allow filters to make logical decisions or calculate elements across time.

There are two basic forms of buffering.

Fixed buffering stores a certain amount (measured in seconds) of data. It's most often used when you want to perform a calculation over a specific amount of time - such as an average/min/max over 5 minutes, for example.

Dynamic buffering will continue to store all information until output occurs.

When output does occur, the buffered data will be flushed for the period up to and including the last datetime index in your output data, if a datetime index exists (otherwise the entire buffer will be erased).

The dedup buffer option prevents duplicate data from being stored in the buffer. This is appropriate for long-term filters that are searching for events, but less useful for filters performing calculations over time.