That’s where Rollup comes into play. The Rollup functionality summarizes old, high-granularity data into a reduced granularity format for long-term storage. By "rolling" the data up into a single summary document, historical data can be compressed greatly compared to the raw data.

For example, consider the system that’s generating 43 million documents every day. The second-by-second data is useful for real-time analysis, but historical analysis looking over ten years of data are likely to be working at a larger interval such as hourly or daily trends.

If we compress the 43 million documents into hourly summaries, we can save vast amounts of space. The Rollup feature automates this process of summarizing historical data.

Details about setting up and configuring Rollup are covered in Create Job API