PGConf.Russia 2018

If you honestly evaluate most of our business applications, you will see that they first collect and import the data into a database and then send the same data in the opposite direction.

What if we don't build an ORM wall between the application and the database, but try using the symbiosis of their strong points and special features instead?

I will tell you how we use PostgreSQL and Clojure for building data-intensive medical applications. We will cover the following topics:

• functional relational programming
• jsonb for modeling complex data domains
• functional indexes and json-knife extension for jsonb search
• graphql implementation on PostgreSQL
• logical replication for building reactive integrations
• asynchronous JDBC-free connector to PostgreSQL on netty

PostgreSQL looks very competitive with other mainstream databases on OLTP workload (execution of large number of simple queries). But on OLAP queries, requiring processing of larger volumes of data, DBMS-es oriented on analytic queries processing can provide an order of magnitude better speed. The following factors limit Postgres OLAP performance:

• Unpacking tuple overhead (tuple_deform)
• Interpretation overhead (Postgres executor has to interpret query execution plan)
• Abstraction penalty (support of abstract data types)
• Pull model overhead (operators are pulling tuples from heap page one-by-one, resulting numerous repeated accesses to the page)
• MVCC overhead (extra per-tuple storage + visibility check cost)

All this issues can be solved using vectorized executor, which proceed bulk of values at once. In this presentation I will show how vector operations can be implemented in Postgres as standard Postgres extension, not affecting Postgres core. The approach is based on introducing special types: tile types, which can be used instead of normal (scalar) types and implement vector operations. Postgres extension mechanism, such as UDT (user-defined type), FDW (foreign data wrappers), executor hooks are used to let users work with vectorized tables almost in the same way as with normal tables. But more than 10 times faster because of vector operations.

How to optimize query processing in PostgreSQL? What if we deal with hundreds of servers and thousands of instances? The company "Tensor" has developed a special tool - explain.sbis.ru, which enables synchronous collection and analysis of queries, visualization of implementation plans, and monitoring of bugs in database.

Pluggable storages is hot subject in PostgreSQL development. The period of heated debates about whether we need them is over. Skepticism about pluggable storages, based on concern that they may be source of inconsistent behavior, was weakened after criticism of PostgreSQL MVCC implementation from Uber side. It became widely understandable that pluggable storages are needed at least for an alternative MVCC implementation. And that is one of way-points for pluggable storages interface design.

At the moment, work on pluggable storages is in the practical stage. There is a thread is pgsql-hackers where few people are developing patchset and several people are doing review.

This talk will cover following subjects:

• overview of pluggable storages interface;
• changes in PostgreSQL core required to implement this interface;
• current and potential implementations of pluggable storages including heap with undo-log and in-memory OLTP engine;
• current state of patchset and prospective of its commit
• further development of interface allowing more possibilities in pluggable storages (columnar, index-organized, LSM and so on).