Elasticsearch Percolate API - Budapest ES Meetup

Transcript

1. Searching backwards
2. So what is this Percolate API? • Think about it as an inverse search or searching backwards. • Usually you store documents and and execute a query to get back the matching subset of the documents. • With the percolator you store the queries and use a document to get back the matching subset of queries.
3. Use Cases • Alerting is a typical use case where your queries are your rules basically. • New item notification on e-commerce sites • Search statistics(details page, listing page, search count)
4. Statistics ~200-300k ads/documents 50+ searchable fields 1M+ searches of interest, ~500k relevant ~100k unique search queries ~200-300k HiFi
5. Demo
6. Scaling • Percolator demonstrates linearly scalability • The memory usage depends on the operators used but in general Percolation is CPU-bound • Filters(metadata fields) and constant scoring • Multi-percolate query • Sharding and routing keys • Dedicated index/cluster
7. Links • Docs • When and How To Percolate - Part 1,Part 2 • Elasticsearch Percolator Continues to Evolve