Speaker Interviews - Steven Simpson

Infrastructure Monitoring with Postgres
Steven Simpson

Devroom, 2017/02/05, 15:00 - 15:50

Could you please introduce yourself? What is your relationship with the PostgreSQL Project? —  I'm primarily a systems level software developer currently working for a start-up called StackHPC. We work on the deployment of OpenStack based HPC clusters. Previous to this, I worked for another start-up called JustOne Database, developing a database geared towards big data which was based on PostgreSQL. During nearly five years in this role, I gained a lot of respect for PostgreSQL and am now hoping to use it in a currently ongoing project.

Monitoring is not something which is a typical DBA job. Which other stakeholders are involved as well? —  The stakeholders in this particular case are system administrators deploying complex computing systems, in particular, OpenStack based HPC clusters. A huge amount of diverse data needs to be collected from these systems in order to ensure smooth operation. We have been evaluating a number of storage systems for the monitoring data which can meet these requirements, and having prior experience with PostgreSQL, decided to evaluate it.

Why are people building complicated infrastructure, if it can be done with one PostgreSQL database? —  Storage of critical data is something you really need a strong understanding of in any system, so being able to learn and trust a single mechanism is still a compelling argument even today, especially for small teams. The database world has undergone significant shifts in the last ten years. There has been a desire to use multiple systems, each of which are built for a specific purpose (such as document storage, time-series, text search or message queues), as opposed to general purpose data storage technologies such as PostgreSQL. This is often in the name of performance, but other reasons are often cited, such as the perceived rigidity of an RMDBS. I feel though this is usually due to the often standoffish relationship between developers and DBAs, more than anything else.

How did you come up with the idea for this talk? —  Well the boring answer is that it is simply a project I am currently working on, but the talk was partly conceived as a means to start a discussion around using PostgreSQL in places for which it is rarely attributed. Suggesting to people that, for some use cases, they could replace five persistent storage mechanisms with one, might garner some interesting discussion.

Have you been to FOSDEM or FOSDEM PGDay before? —  This is my first FOSDEM, though I have visited PGDay previously in the UK, and will look to return if the opportunity arises.