Today, we want to congratulate Prometheus on its tenth birthday!
From Google to SoundCloud
Ten years ago today, the first public commit was pushed to the Prometheus repo. Back then, Matt T. Proud and myself had just left Google to join SoundCloud. At SoundCloud, we encountered an environment that was ahead of its time: Hundreds of microservices were running on an container orchestration system that SoundCloud had built in-house, before Kubernetes or even Docker were born. However, SoundCloud was also facing a myriad of reliability and latency issues, and existing monitoring tools had a hard time making sense of such a dynamic and high-cardinality computing environment. We felt unable to pinpoint the sources of problems, and we were really missing a system that combined a dimensional data model with a flexible query language, an efficient TSDB implementation for tracking dynamic high-cardinality time series data, service discovery support, as well as integrated alerting. All these were qualities that we had gotten used to from Borgmon, Google's own internal monitoring system at the time.
We felt so strongly about the need for a new open-source monitoring system that we decided to take a stab at building it ourselves. Without even asking anyone at the company, we threw some code on GitHub under an open license and started working on it in our free time (with that "free time" gradually turning into more and more time on the job). Matt focussed on building client instrumentation libraries and a TSDB implementation, while I built the initial PromQL query language, the web UI, as well as the configuration handling. Toward the end of 2012, we had a rough prototype that could scrape metrics from an instrumented service target, store them in a TSDB, and then query and graph those metrics using an early form of PromQL. This end-to-end proof of concept gave us the confidence to keep going: Prometheus was born.
The journey since then
Over time, and after overcoming a lot of technical and political hurdles, SoundCloud adopted Prometheus as its primary monitoring system, which gave developers unprecedented insight into their service behaviors. We also finally announced the project properly to the wider world in 2015. The rest is history: Prometheus made it to #1 on Hacker News and hit a nerve in a world where early Kubernetes users were looking for a monitoring system to go along with the shift to more dynamic computing environments. Ex-Googlers also immediately spotted the uncanny resemblance of Prometheus to Borgmon and were happy to find similar technology outside of Google. Adoption skyrocketed, the Prometheus team grew over the years, and the project joined the CNCF as the second project after Kubernetes. Prometheus released a 1.0 and then a 2.0 version, became mature, efficient, and feature-rich, and has now become the de-facto standard for metrics-based instrumentation in the world. Even the large cloud companies such as Amazon, Google, and Microsoft now offer managed Prometheus-style monitoring services, and Prometheus interfaces for data collection, querying, and more can be found implemented everywhere. Overall, Prometheus has enabled an era of metrics-driven development, monitoring, and alerting, that allows for much better visibility into systems and services than solutions before it.
This journey would not have been possible without a lot of great people and organizations: A special thank you goes out to the excellent and ever-growing group of Prometheus Team members that have turned Prometheus into what it is today, to the wonderful user community who are using, testing, and spreading Prometheus around the world, as well as to the CNCF and a whole ecosystem of companies (of which my company PromLabs is just one) that are pushing Prometheus-based monitoring forward in new ways. Recently, the team at Speakeasy Productions also produced the Inside Prometheus film, a documentary about the history of Prometheus, which I recommend watching.
Finally: If you are looking to learn Prometheus from the ground up or train your engineering team, have a look at our self-paced Prometheus training courses, authored by myself as the co-founder of Prometheus.
Comments powered by Talkyard.