Presentation: Etsy Search: How we index and query 26 million one-of-a-kind items

In this talk talk I pull back the covers on the Etsy Search ecosystem and how we got here -- the good, the bad, and the funky.

The heart of our system is patched Solr/Lucene & ElasticSearch (and a lot of Bash) running on our boxes. I’ll dig into: signal collection from our web stack and mobile apps into Hadoop, generating machine learning ranking signals with Scalding, indexing content from 900 MySQL shards, sharding our Solr index in a novel way for performance, building a native search advertising platform, a 4-byte silver bullet for fighting JVM GC pauses, and how we continuously deploy this whole thing. I'll also talk about a bunch of things we tried that didn't work. 

The goal of all this is to connect people: So our 40 million members can find unique goods that inspire them from our ever-changing set of one-of-a-kind items produced by a million individual sellers and makers around the world.

Aaron Gardner Elsewhere

Similar Talks

Tracks

Covering innovative topics

Monday, 3 November

Tuesday, 4 November

Wednesday, 5 November

Conference for Professional Software Developers