We added new metrics and functionality to the console. You can now see a breakdown of which system processes use memory for a Virtual Instance and a new chart that shows the Cache Hit Rate. There is also a new time picker functionality and a new crosshair sync behavior between all the charts on the same page that shows the metric data for that point in time for the chart that is being hovered over.

You can now add sources to and remove sources from a collection in the console. Learn more about adding and removing sources here.

You can now split the results of a INSERT INTO s3 query such that records with the same field value will be emitted together in files prefixed with the value. Learn more here.

The ability to export the query results from Rockset and write them directly to Amazon S3 in the Parquet data format. Learn more here.

Used to calculate the edit distance between two strings s1 and s2 according to the algorithm specified. This new EDIT_DISTANCE function replaces UDFs, offering up to a 10x speed-up on fuzzy search queries. Learn more here.

Rockset now supports Parquet files using ZSTD compression as an ingestion format, making it easier to ingest compressed data for search and analytics. Learn more here.

Vectorized wide integer types to more efficiently process numeric and decimal values that need to maintain precision. We recorded a 10x improvement in performance with these optimizations.

ARRAY_CONTAINS_PREFIX(string_array, prefix) returns true if any element of string_array starts with the given prefix.This new function Improves the ergonomics of prefix search and results in better performance on nested arrays. Learn more here.

Vectorized MIN and MAX operations for timestamps and dates for better performance of time series analytics. Internal benchmarks show a performance improvement of 82% with vectorization.

Decreasing the number of memory allocations across sequential hash computation operations, thus improving efficiency of SQL joins.