Red Sky Ops is composed of three parts:
- The Red Sky Ops Manager, which runs on your cluster,
redskyctl, a CLI tool for interacting with the manager, and
- An optional server (available as an enterprise product) that automatically generates suggested configurations for your application.
The Red Sky Ops Manager is a Kubernetes operator written in Go. It is comprised of several Custom Resource Definitions (CRDs) and a Controller that runs in your cluster in a dedicated namespace. The CRDs allow you to create Experiment and Trial objects in your cluster. The controller manages those experiments and trials, and communicates with the Red Sky Ops Server if configured to do so. The manager can be installed on your cluster using
redskyctl init (see Installation).
The Red Sky Ops Tool
redskyctl is a command-line tool for interacting with the manager. It can be used to install the manager, create trials by assigning parameters interactively, and more. It is meant to be used in combination with
If you are on a Red Sky Ops Enterprise plan, you can configure your cluster to connect to a Red Sky Ops Server. The server produces experiment suggestions automatically.
An Experiment is the basic unit of organization in Red Sky Ops. The purpose of an experiment is to try configurations of an application and measure their impact. To accomplish this, experiments:
- Make applications manifests configurable via parameters
- Run trials with specific values assigned to each parameter
- Assess the outcome of each trial by measuring one or more metrics
Parameters are the inputs to an experiment. A parameter has a minimum value, a maximum value, and a name. Currently, only integer parameters are supported. Taken together, an experiment’s parameters define the search space: the total domain of all possible configurations. For in-depth explanation, see the parameters page.
Trials are single runs of an experiment, with values assigned to every parameter. An experiment typically consists of many trials.
Metrics are the output or outcome of a trial. They are used to measure the result of a particular choice of parameters. A metric is a numeric value that an experiment attempts to minimize (like cost in dollars) or maximize (like throughput) by adjusting the values of parameters. For in-depth explanation, see the metrics page.