Case Study: Wikimedia Foundation achieves scalable, real-time multilingual search across 888 wikis with Elastic

Wikimedia Foundation - Customer Case Study

Wikimedia’s search team needed to power discovery across a massive, multilingual ecosystem — roughly 888 wikis in about 265 languages — serving editors and readers worldwide. The challenge was extreme scale and diversity: in December 2014 the system handled ~870 million full-text searches (and billions more prefix searches), indexed ~200 million documents across >1,800 indices (~6,500 shards), and had to support communities with varying language needs while providing near–real-time updates and editor-facing query tools.

They built CirrusSearch on Elasticsearch, contributing plugins and engine improvements (highlighting, phrase suggester, trigram-accelerated regex, multipass rescore) and exposing advanced operators for editors. The deployment delivers near–real-time indexing (most edits reflected in under a minute), runs on a compact fleet (31 servers with 2x replicas), and scales to hundreds of millions–to–billions of queries while enabling further work on relevance instrumentation, multilingual search, mobile integration, and better Wikidata search.

Wikimedia Foundation

Chad Horohoe

Senior Software Engineer