- ElasticSearch: With the same amount of data ingested into Elasticsearch and Pinot, Elasitcsearch’s memory usage was 4x higher and disk usage was 8x higher than Pinot. In addition, Elasticsearch’s query latency was 2x-4x higher than Pinot, benchmarked with a combination of filters, aggregation and group by/order by queries.
- Apache Druid: Pinot is similar in architecture to Apache Druid but has incorporated optimized data structures such as bit compressed forward indices, for lowering the data footprint. It also uses specialized indices for faster query execution such as Startree [39], sorted and range indices, which could result in order of magnitude difference of query latency. Recent studies done outside of Uber also confirmed this performance benefit of Pinot over Druid [47] [61].
https://arxiv.org/pdf/2104.00087.pdf