Druid Feb 2017 Roadmap

Druid 2017 Roadmap from Feb 2017 meetup

Transcript

1. Druid 0.10.0… and beyond!!
2. 0.10.0 - headliners • built-in SQL (cli, http, jdbc) • group/filter on numeric columns without cast to string • optional kerberos auth between druid nodes • performance improvements • optimized large amounts of and/ors with concise bitmaps • index-based like ‘foo%’ filter • short-circuiting and filter • ~30% improvement on non-time groupBys • 2x improved cross-node transfer rate for queries with large resultsets • 0.10.0-rc1 due soon (8/236 issues still open) • requires java 8
3. …and beyond!! • reindexing / compaction without hadoop • improved schema definition & management • closer hive/druid integration • performance improvements • jit-friendly topN • limit push-down for groupBy • security layer (row-level / column-level) • work towards joins
4. built-in sql • like hive integration, also based on Calcite • all queries are executed as native Druid queries • overhead is minimal • choose between approximations or exact results • includes druid extensions like APPROX_COUNT_DISTINCT, APPROX_QUANTILE • some sql features are unsupported – working to reduce this footprint over time