Materialized Pre-Aggregations for Exploratory Analytics Queries

Analytics queries often access large numbers of documents where query performance can suffer due to high disk I/O. If the aggregation pipelines are known in advance, materialized views can be created to speed up performance. But what if the queries are ad-hoc and not known ahead of time? In this talk, developers and data scientists familiar with MQL will learn a technique to pre-aggregate data without losing the flexibility for ad-hoc querying. The method is related to the concept of OLAP cubes in relational systems, but applied to MongoDB and using the aggregation framework. We discuss the approach, show example pipelines, and go through the limitations of this technique.