Manticore Search has introduced new aggregation functions for percentiles, percentile ranks, and median absolute deviation (MAD). These tools allow users to analyze data distributions, outliers, and latency tails directly within search queries.
The new functions include percentiles, percentile_ranks, and median_absolute_deviation. Users can control the balance between memory usage and calculation accuracy using the optional compression parameter, which defaults to 200.
The system supports both SQL and JSON API syntax for these aggregations. These features are designed for use cases such as monitoring latency, analyzing e-commerce pricing, detecting anomalies, and reporting on service level objectives.
The implementation uses approximate algorithms to maintain predictable performance and memory consumption. Developers are advised to start with the default compression setting and adjust it if higher precision is required.
