MiCADO - autoscaling framework for Kubernetes Deployments in the Cloud

This software is developed by the COLA project.


MiCADO is an auto-scaling framework for Docker containers, orchestrated by Kubernetes. It supports autoscaling at two levels. At virtual machine (VM) level, a built-in Kubernetes cluster is dynamically extended or reduced by adding/removing cloud virtual machines. At Kubernetes level, the number of replicas tied to a specific Kubernetes Deployment can be increased/decreased.

MiCADO requires a TOSCA based Application Description to be submitted containing three sections: 1) the definition of the individual applications making up a Kubernetes Deployment, 2) the specification of the virtual machine and 3) the implementation of scaling policy for both scaling levels. The format of the Application Description for MiCADO is detailed later.

To use MiCADO, first the MiCADO core services must be deployed on a virtual machine (called MiCADO Master) by an Ansible playbook. MiCADO Master is configured as the Kubernetes Master Node and has installed the Docker Engine, Occopus (to scale VMs), Prometheus (for monitoring), Policy Keeper (to perform decision on scaling) and Submitter (to provide submission endpoint) microservices to realize the autoscaling control loops. During operation MiCADO workers (realised on new VMs) are instantiated on demand which deploy Prometheus Node Exporter, CAdvisor and Docker engine through contextualisation. The newly instantiated MiCADO workers join the Kubernetes cluster managed by the MiCADO Master.

In the current release, the status of the system can be inspected through the following ways: REST API provides interface for submission, update and list functionalities over applications. Dashboard provides three graphical view to inspect the VMs and Kubernetes Deployments. They are the Kubernetes Dashboard, Grafana and Prometheus. Finally, advanced users may find the logs of the MiCADO core services useful on MiCADO master.