lkafle / iter8

Kubernetes release optimizer

Iter8: Kubernetes Release Engineering made Simple

What is Iter8?

Kubernetes release engineering toolkit built for DevOps, MLOps, SRE and data science teams.

What is an Iter8 experiment?

Iter8 experiments make it simple to collect performance and business metrics for Kubernetes apps and ML models, assess and compare multiple app/ML model versions, safely rollout winning versions, and maximize business value with each release.

Example

The following picture illustrates an Iter8 experiment that performs load testing with SLO validation of a gRPC service.

Experiment chart

In order to enable reuse, Iter8 experiments are templated and packaged as specialized Helm charts. Experiment charts can be combined with values to generate experiments.yaml files that provide fully defined experiment specifications.

Iter8 experiment charts enable to you to launch powerful release engineering experiments in a matter of seconds.

Features at a glance

• Load testing with SLOs

  Iter8 experiments can generate requests for HTTP and gRPC services, collect built-in latency and error-related metrics, and validate SLOs.

• A/B(/n) testing

  Grow your business with every release. Iter8 experiments can compare multiple versions based on business value and promote a winner.

• Simple to use

  Get started with Iter8 in seconds using pre-packaged experiment charts. Run Iter8 experiments locally, in a container, inside Kubernetes, or inside your CI/CD/GitOps pipelines.

• Awesome integrations

  Use with any app, serverless, or ML framework. Iter8 works with Kubernetes deployments, statefulsets, Knative services, KServe/Seldon ML deployments, or custom Kubernetes resource types.

Installation

Install the latest stable release of the Iter8 CLI using brew as follows.

brew tap iter8-tools/iter8
brew install iter8

You can also install Iter8 using:

• pre-compiled binaries
• source
• go 1.16+

Usage Examples

HTTP load test with SLOs

1. Load test an HTTP service and validate SLOs.
2. Control the load characteristics during the HTTP load test experiment.
3. Load test an HTTP POST endpoint with request payload.
4. Learn more about built-in metrics and SLOs in an HTTP load test experiment.
5. Load test a Knative HTTP service.

Documentation

Iter8 documentation is available at https://iter8.tools.

Contributing

See here for information about ways to contribute, Iter8 community meetings, finding an issue, asking for help, pull-request lifecycle, and more.