DevOps companies are in the business of observability. The sooner they realize that, the better.
Observability has come to mean more than just logs, metrics, and traces. It has encompassed the entire lifecycle of system management – from alerting on anomalous data to securing endpoints. Observability helps businesses manage their software.
What does that have to do with DevOps? At a high level, DevOps products sit in the software development lifecycle – from development, to CI, to deployment. It also includes all of the glue that keeps everything together, the building blocks that higher level business applications use (e.g., a DAG executor, a secrets manager, an event streaming platform).
What are DevOps companies selling?
- Not the hardware. Many customers elect to run these services on-premise or cloud-prem for data security. The vendors themselves run on cloud, so they don't have any competitive advantage in selling raw compute or storage.
- Increasingly, not the software. Many DevOps companies use open source as a go-to-market strategy. You can run Vault from HashiCorp yourself for free. Gitlab is free to manage yourself. Even JFrog, a closed source platform, could be put together with open source components.
- Not the installation. In the past, there were systems integrators who specialized in helping you connect all of your software appliances together. Now, it's as simple as deploying a Helm chart on your Kubernetes cluster, or using any of the 20 services that run containers on AWS.
- Not professional services. While many of these companies will offer professional services like training, it isn't the source of scalable recurring revenue that makes these companies venture fundable.
What's left is managing the software – keeping the service running. As creators of the software, these vendors are experts in running it. To do this efficiently, the vendors create their own observability tools – managed control planes, performance optimizations, customer dashboards, and higher-level metrics APIs.
But observability-as-a-service isn't enough for these managed SaaS platforms. That's because no DevOps SaaS exists in a vacuum. Any service on the critical path needs to export observability metrics somehow. Some of this comes for free with on/cloud-prem software (profiling, infra monitoring). But to debug upstream (or downstream) issues, engineers need to collect observability logs, metrics, and traces in the critical path end-to-end.
So DevOps companies should think about how they are positioned as an observability provider. Good management comes from good observation.