PGDay London
Russell Morris
- 8 minutes read - 1578 wordsToday I went along to the PG Day London. After visiting similar conferences in the past I was on the lookout for some interesting talks and refreshing my gardening wardrobe (my partner is not keen on being out with me in public wearing database software branded t-shirts).
So after braving the walk up from London Waterloo to where the conference was held (thanks TFL and the unions). I was looking forward to a cup of coffee, maybe a pastry and the keynote speech (I made it just in time after).
Sadly the coffee disappointed but that was the only thing that did, here we go with my summary of the day - which I hope you will find interesting. If you would like to avoid my opinions please head direct to link for the presenters/slides and other magic.
Future Postgres Challenges - Bruce Momjian - EDB
So in this talk Bruce went through the current position of Postgres and also possible future problems that could be on the horizon - he came up with several interesting stories, but also reasons for why Postgres is different from other open source offerings and hopefully will be safe from future ill winds.
- Development and leadership of Postgres are globally diversified which should hopefully mean its safe from the ill tempers of governments etc
- There are a large number of contributors (edb, aws, microsoft, ibm….) all contributing to the project
- Small chance of one contributor taking control of the project as its part of the setup restricting this happening
- As there’s so many contributors it means that releases are wide and encompassing rather one contributor controlling direction
- Hopefully relatively safe from patent challenges and it’s reputation is good (so far)
- Relational databases seem to be here to stay - there have been alternatives but they all have a place
- Postgres is flexible enough to adopt components that work say in JSONB without having to change direction entirely
There are still technical challenges with Postgres:
- Write amplification on insert/update with indices and maintenance overhead
- Cluster file encryption (TDE)
- He should have added row level security (as seen in later talk)
- Horizontal scaling - progress is being made but there are issues here
Bruce seemed to take the example of lets show the bad and the good of PostgreSQL to the community and we can all address these.
Postgres schema and migrations using the expand/contract pattern - Andrew Farries - Xata.io
Ok I may have skipped this one as its the same talk I had earlier in the PostgreSQL user forum…I will copy from the schedule, their summary.
Database schema migrations often pose significant challenges to developers, particularly when striving for zero downtime migrations.
There are many things that can go wrong:
- Schema changes breaking client applications
- Unexpected table locking causing downtime
- Human mistakes causing data loss
This talk presents a new approach to schema migrations using an ‘expand/contract’ pattern where multiple versions of a database schema are maintained during the migration, allowing old and new versions of client applications to run side-by-side during an application rollout. Each version of the application sees the version of the database schema with which it is designed to work.
The talks covers the ideas behind this approach, and the challenges to be overcome such as:
- Limitations of existing tools and techniques for Postgres schema migrations.
- The use of Postgres views to present multiple versions of an underlying table.
- Backfilling data between old and new schema versions.
- Techniques to avoid unexpected table locking during migrations.
Finally we introduce pgroll, an open-source tool that puts these ideas into practice and we show how it can be used to facilitate safe application rollouts across database schema changes.
PostgeSQL Performance Tuning - Micheal Banck - credativ GmbH
Micheal was very interesting in this talk to be honest. I have copious pages of notes about possible changes to implement.
The talk focused on key parameter setting changes in PostgreSQL that can then give positive optimisation improvements to your systems. It was a nice change seeing the changes described with the pitfalls rather then reading the changes from blogs. You need to remember that a lot of the settings in PostgreSQL were set a number of years ago and technology has moved forward but sadly the default settings have not.
Please go and read his slides - you may learn something (I did).
Serving Literary Knowledge as Data : Building and Documenting DH Apis with PostgreSQL - Marta Palandri
Now this would have made my partners day - she is currently writing her dissertation for her Masters degree and this would have made her life so much easier.
So in this talk Marta went though taking study text and how you can use Postgres and API’s to be able to interpret it in a meaningful manner. I have included her summary below because, although the talk was very interesting - its not something I feel I can adequately summarise
In Digital Humanities (DH), the challenge of making interpretive research—such as literary symbolism, historical analysis, or cultural studies—queryable and accessible can be met through powerful APIs built on relational databases. In this talk, we’ll demonstrate how PostgreSQL can serve as the backbone for creating an API that transforms subjective analysis into structured and queryable data.
Using examples from symbolic analysis of cultural texts, we’ll walk through key PostgreSQL features, like jsonb, views, and full-text search, that enable the modeling, querying, and retrieval of interpretive knowledge.
Additionally, we’ll cover best practices for generating API documentation using tools such as Swagger and MkDocs, ensuring that your APIs are accessible not just to developers, but also to scholars who may not be familiar with the underlying data models. The session will also address the epistemological implications of exposing interpretive data as a public dataset: How to manage uncertainty, signal multiple interpretations, and what it means to present human analysis in a structured and queryable format.
Viewers will gain practical insights into using PostgreSQL to support Digital Humanities projects and learn how to create and manage APIs that make interpretive knowledge more accessible to a broader audience.
Explaining PostgreSQL like I’m Five (or 85) - Ellyne Phneah - DBTune
Now this was a topic we can all get on board with. When your mum asks what you have done today - how do you explain that you have created a 10GB index on a database table to improve query speed- without eyes glazing over and being offered a slice of cake and a cup of tea.
Ellyne had this issue when describing what she did to here tech savvy granddad. So this talk revolved around being able to communicate those techy pieces of data in a manner thats understandable to all users using metaphors and story telling arcs.
I know that I am going to use some of the examples going forward.
PostgreSQL certification - The PostgreSQL way - Jan Karremans - CYBERTEC
Now I was really looking forward to this. A lot of databases have recognised certification paths and PostgreSQL has many admin courses, but nothing really industry recognised
So the idea is that this is by the community - for the community. A recognised certification that proves that you have your Postgres spurs. Whilst its still in its early days you can already see the opposition that this may encounter. I like the idea - and will monitor the progress of the certification with interest, who knows one day I may be certificated and not just to prove sanity.
Adaptive Partitioning : Balancing OLTP and OLAP with Citrus
I am sorry post lunch and lack of caffeine - as no breaks for the last 2 sessions caused my attention levels to droop on this one, I have included the summary from the site below as it was interesting but for me I am not overly sold on this.
Modern applications often need to handle a blend of transactional (OLTP) and analytical (OLAP) queries on the same dataset—posing a unique challenge for traditional database architectures. Citus’s sharding and partitioning features can help strike that balance. In this session, we’ll cover how to design a table layout that gracefully serves high-velocity writes and transactional queries while still enabling fast analytical aggregates. We’ll look into practical partitioning patterns, index strategies, and how to leverage co-location for analytical joins. We’ll also discuss how to monitor workload patterns so you can adjust partition sizes and distribution keys as data or query demands evolve. After this talk, you’ll have a roadmap for building adaptive data models in Postgres with Citus that can fluidly handle both OLTP and OLAP workloads.
Row-Level Security sucks. Can we make it usable? - Jimmy Angelakos
Now Jimmy introduced this as the most boring talk of the day. To be fair he was fair from the truth.
What we were shown was a balanced view of the Row Level security model
- how it should be implemented
- what the pain points of the implementation are
- how these can be mitigated
- the question of do you really want to do this
- a solution of how it can be implemented in a secure and working method
To be fair it was a great talk - as something with so many caveats and work arounds I am not sure I am going to use it just yet
Overall this was a good day, the sponsor stands were interesting - lunch was very tasty and good talks. Looking forward to next year already.