PgConf.Russia 2017

In the data persistence layer, using PostgreSQL from the very start of development, we went all the way from a small cluster on a virtual machine to a multi-host system that provides near real-time processing of mixed OLTP/OLAP load. In this talk, I’m going to tell you about the main development stages of our analytical solution at the application and infrastructure levels, and describe the specifics of using PG that we encountered.

VIDEO

I will tell you about why Postgres is first-choice product as a foundation for your BI system with classical OLAP workload. Briefly it will be said about existing open source BI solutions.

I will also describe specific of our architecture, why we chose snowflake scheme and how we are doing extract, transformation and load procedures. It will be mentioned about special Postgres tuning for OLAP and massive data bulkload workloads. Also I will let you know about Postgres usage as a column database with cstore_fdw by Citus and results achieved. Cons and problems of our approach will be described in the end of the talk.

VIDEO

When working with a complex business logic, you often have to implement a workflow - a sequence of several processing steps, with each step implementing a separate part of the business logic. This is usually done with specialized queues, but if there are high reliability demands, it makes sense to do everything on PostgreSQL.

I will describe the tasks that require a workflow implementation, offer a solution, compare it with other options, and tell you about the implementation traps and pitfalls.

Citus allows you to distribute postgres tables across many servers. It extends postgres to transparently delegate or parallelise work across a set of worker nodes, enabling you to scale out the CPU and memory available for queries.

One year ago, we began a long journey to allow Citus to scale out another dimension: write throughput. With writes being routed through a single postgres node, write throughput in Citus was ultimately bottlenecked on the CPUs of a single node. Citus MX is a new edition of Citus which allows distributed tables to be used from from any of the nodes, enabling NoSQL-like write-scalability.