cognative
cognative is an opinionated, minimalistic approach to business intelligence 🧠 and operations 🚨.
Goals
- Reduce the complexity and number of systems operations staff is required to have expertise on.
- Maximize the breadth of functionality they are able to support across business intelligence and operations.
- Simplify the experience for developers.
- Enable companies to own their own data.
- Be flexible enough to support a variety of deployment models.
Background
I started cognative as a recent reflection on the operational constraints faced by smaller organizations and how to make best use of their skills and expertise. Requiring operations staff to be experts on a large number of systems is not sustainable and will only lead to burn out. And yet, an increasing number of complex systems need to be run to support business intelligence and operations.
The project name is a clever play on words. It takes the acronym for the tech stack (COG) and joins it with “native”. It came out of a discussion I had with Tim Banks about the pitfalls of “traditional” operations tech stacks and how it’s often short-sighted once you start to consider the broader needs of business intelligence. COG stands for Clickhouse, OpenTelemetry, and Grafana.
- Clickhouse is our data warehouse layer. It stores logs, metrics, traces, as well as other business related data sets. Clickhouse also comes with tons of integrations which makes it easy to source data from a variety of locations.
- OpenTelemetry provides instrumentation for our ecosystem. Information is collected from a variety of sources using vendor neutral solutions. The community also supports a large number of languages, making it easy to add to many systems today.
- Grafana provides our data exploration, visualization, and alerting layer. It can easily integrate with on-call solutions like PagerDuty, OpsGenie, BetterStack, and many more.
Helm
You can easily deploy the cognative stack to Kubernetes using Helm.
helm repo add mya https://mya.sh
The current version of the Helm chart only supports a single instance of Clickhouse. It is great for testing out the
stack, and smaller scaled deployments that don’t require high availability. For a more critical, production-based
workload, we recommend disabling the clickhouse portion of this chart, and leveraging Clickhouse Cloud in the
meantime. You will also need to configure your connection details to have the deployment connect to your managed
instance.
helm upgrade -i cognative mya/cognative