Full-text search X times faster: Inverted index vs. SQL OLAP

Inverted indexes, popularized by Elasticsearch and Lucene, were designed for consumer web search with rich tokenization and scoring. For observability and security log workloads, these features become antipatterns — developers need grep-style substring matches, not stemmed token lookups. SQL OLAP databases like ClickHouse can be 10x+ faster and more cost-efficient for this use case by using raw scans with skip indexes (e.g., bloom filters), columnar storage with tailored compression, and vectorized query execution. The post also discusses the migration challenge: organizations like Uber have 10k Kibana dashboards they can't easily move, while others like Zomato saved $1M/year by switching to ClickHouse. The author's product, Quesma, aims to bridge the Elastic ecosystem with SQL analytical backends.