Full-text search: from scratch to a HA cluster
Search is easy. Just how hard can a CREATE FULLTEXT INDEX can be, right? Except that won't scale. And it won't be easy to tweak either. And don't even get me started on relevance. (And no, relevance isn't something only Google can have, because they enslaved every scientist working on that.)
Proper search at a decent scale is tough. Whatever technology you choose. Getting it right will require time and effort. There's no silver bullet. We give you a lead bullet though: Sphinx. And it comes with a silver lining: this tutorial. In which we shall go through several different topics and demos:
- Getting started with Sphinx. Using SphinxQL and SphinxAPI. And a 20-minute integration demo.
- Managing indexes. Disk indexes, delta scheme. Realtime indexes.
- Search quality. What is relevance. Judgments, metrics, formulas, experiments. What else is there to search quality.
- Cluster setups. Distributed indexes. Capacity planning. HA tools in Sphinx.
So come drop by if your current search "solution" sucks and is more of a problem than a solution. We know a few tricks to make it better and we want to share.