[keynote speech] Schema-less PostgreSQL

Oleg Bartunov
Moscow State University (MSU) Sternberg Astronomical Institute / PostgreSQL Developer

MongoDB reigns in the world of semi-structured data. Attracted to MongoDB investments often overshadow the mind of developers (especially naive beginners), who are happy to throw themselves into the ocean of opportunities offered by NoSQL (it”s cool!).
Enthusiasm fades after realizing the fact, that nothing is free, and one should think about functionality, that already works fine for decades in traditional relational databases, which do an excellent job with loads and data for 99% projects.

For more than 10 years PostgreSQL provides the opportunity to work with schema-less data.
One of the most widely used extension hstore offers a “key-value” storage, while retaining all the power of relational database.
However, an emerging technologies, based on the popular standard JSON, require built-in database support for hierarchical structures.
Document-oriended NoSQL databases uses JSON format to aggregate data into documents and have arisen from the need to adapt fast to new customer needs,with instant schema changes, whereas schema migration has always been a problem of relational databases.
PostgreSQL since 9.2 release has JSON data type, which was not really useful because of textual storage and poor operations support.
Upcoming 9.4 release provides new JSONB data type with binary storage and fast operators with indexes support.
PostgreSQL, being a true relational database, outperforms MongoDB, because of better indexes, which are faster and smaller.
JSONB provides PostgreSQL users instant schema migration along with traditionally strong ACID support and powerful SQL.

I will talk about technologies behind of functionality and performance of the new JSONB data type and present our new extension JSQUERY, which provides rich JSONB query language with indexes support.

I graduated with honour the Moscow State University (MSU) in 1982 and since then I work in the Sternberg Astronomical Institute, Moscow University as a research scientist.
I started working with PostgreSQL since 1996 and my main contributions are locale support, indexing infrastructure GiST, GIN and SP-GiST, full text search system, data types for efficient work with semi-structured data (hstore, jsonb) and extensions like intarray, pg_trgm, ltree and some others. Currently, I’m working on better indexing system (VODKA) and jsonb improvements.

I like to play volleyball, trekking in mountains, running and making photo.
Sometimes I write posts in my personal blog at obartunov.livejournal.com and publish pictures at www.flickr.com/photos/obartunov/.

Also, you can find me at www.facebook.com/oleg.bartunov.