Automated Deadline-Based Scaling of Experiments in the Cloud with MiCADO

James DesLauriers

Many scientific applications require computation resources based on dynamically changing requirements. The H2020 COLA (project-cola.eu) set out to solve this with a generic framework (MiCADO) to support automated scalability of applications deployed to the cloud. MiCADO is open-source, cloud agnostic and uses widely applied technologies - Kubernetes, Occopus and Prometheus - to scale virtual machines and containers in the cloud. MiCADO has been tested on a range of small to large-scale research activities that saw simulations and other multi-job experiments queued, executed and scaled in containers on different public and private clouds. The talk will focus on MiCADO, its architecture, characteristics, and its benefits to the RSE community.

🎥 This talk was recorded on video and is available at https://doi.org/10.5446/42564.

🖥 There is a PDF file with the slides for this talk available. Unless otherwise noted in the slides themselves, they are published under a CC BY 3.0 license.