Sphinx 2013
summary
Sphinx 2.1/2.2 adds concrete, usable features: built‑in adaptive HA and replica load shifting, full JSON document storage with real‑time insert/replace and a 4GB internal blob limit, and attach/alter/optimize to merge fast batch disk indexes into RT indexes. Morphology gets Arabic stemmer and lemmatization modes, plus token splitting for domains/compounds. New ranking exportable ranking signals and a PTF-style factor allow ML training; skip lists and bigrams accelerate queries (skip lists ≈+70% on some tests, bigrams up to 10–100× for phrases) while bigrams enlarge index ~15–30%. Profiling and execution‑time prediction tools ship and work.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 7:05High availability and load balancing
- 10:19Search quality and morphology
- 20:55Ranking and signals
- 30:10JSON and attribute support
- 37:55Dynamic indexes
- 40:25Special features
- 44:43Instrumentation and profiling
- 48:35Optimizations and performance
- 56:17Q&A
related talks
57mUnlocking the Power of Hybrid Search: A Deep Dive into Python-Powered Precision
40mWhat's evolving in Elasticsearch and Lucene?
39mSearch Architecture at Avito
51mHibernate Search talk
28mA Deep Dive into Python-Powered Precision and Scalability
33mHydra - An Open Source Processing Framework
51mSix New Important RocksDB Features And Planned Works
40mKhepri - Mnesia replacement