Boriss is a Solution Architect at EDB. He’s got his PhD researching distributed self-managing systems and he has been recognized as a PostgreSQL contributor. He started the PostgreSQL User Group in Belgium in 2018 and is one of the organizers of PGDay Lowlands. He has been a presenter at many Conferences in Academia, Open Source, and Postgres, and has presented a few times already at PgConf Belgium. He’s also the co-author of the “PostgreSQL 16 Administration Cookbook”, along with Simon Riggs and others. In his free time, he performs with his band Trasandinos, bringing music to connect people. He’s also an air guitar player and loves to spend time with his family.
When vertical scalability isn’t enough to bring your application to the next level, software architects start thinking about scaling horizontally, also known as scaling-out. While this approach works very well for stateless components, mostly at the front-end of the application, scaling out the database, the stateful part of your application, is a totally different challenge. Spoiler alert: it can’t be done transparently. Therefore, awareness of what needs to be taken into account is valuable information for software developers and architects.
In this presentation we will analyse advantages and challenges when using active-active databases (also known as multi-master), creating awareness of the implications of conflict resolution, consensus, sharding, and other concepts that are intrinsic to distributed systems. To make the presentation more concrete, we will use EDB Postgres Distributed to see how many of these challenges have been solved, and how they are used in production.
We know this will sound cliché, but there is a bit of truth in saying that it is not difficult to spot a DBA, because they are always tuning their beloved databases, talk about IOPS performance, and nowadays, they complain about their databases being moved to Kubernetes. On the other hand, Kubernetes folks want to turn every project, however simple, into a microservices system, and they answer nearly every question with “I could write an operator to solve that”. One could say that their fanaticism makes them see Cloud Native almost as a religion.
In this presentation, we don’t want to criticize either of these two profiles, nor their approach to problem-solving and life in general. We simply cannot deny the fact that many companies are turning into Kubernetes to manage their entire infrastructure. Therefore, this talk is an attempt to help DBAs to change their mindset when working with databases running on Kubernetes. We don’t want them to be seen as sinners and face the Cloud Native Inquisition.
To make the ideas more concrete, in this presentation we will focus on examples running CloudNativePG, which is now part of the Cloud Native Computing Foundation and is the most popular operator to run PostgreSQL. The ideas are valid for other databases too.