In industry, containers and container orchestration solutions have dramatically changed the way system administrators deploy and manage applications. Developers are gradually switching from delivering monolithic applications to micro-services.
Indeed, using containerisation solutions such as Docker has many advantages: applications run in an isolated way, decoupled from the operating system and its libraries; runtime dependencies including access to persistent storage are clearly declared.
Container orchestration, as provided by Kubernetes or OpenShift, allows for an even higher level of abstraction: for an application to run on a computing infrastructure, only a declarative statement containing runtime parameters is required. These parameters include the number of instances or the maximum amount of volatile memory to be allocated. Thus, complexity is hidden and scalability can be easily achieved.
However, switching to containers has some drawbacks: firstly, it requires an initial investment in the computing infrastructure, training and new disaster recovery procedures. Secondly, additional software layers create operational risks, and potential security threats. Finally, as the ecosystem is new, frequent incremental changes need to be applied on the infrastructure itself, which may be problematic in the context of accelerator controls.
This workshop aims to share knowledge and experience on these technologies. We will explore what are the use cases in laboratories and large experiments, how to make the best of them, and their limits.
Registration
Registration and accommodation are handled through the ICALEPCS 2019 website.
Please register here: https://icalepcs2019.bnl.gov/ before submitting your abstract on this Indico page.
