Unikernel based deployments
Submitted by alok singh (@alephnull) on Friday, 27 March 2015
This session is meant to show how applications can be built for unikernels and show an actual deployment scenario.
Deployment engineering for things that need to run at scale seems to be based on building images a la Docker et al. While this is an improvement over the current state of affairs, there are operational challenges in implementing such a scheme such as image size, infra required to build the images, a large surface area of attack. Unikernels such as MirageOS provide a way of addressing some of the challenges.
We will take a WSGI application and unikernalise it while trying to come up with a workflow that could be generic enough to apply to WSGI applications.
We will then deploy this on a toy setup and discuss the deployment challenges and sources of the challenges. For instance, scale is one of the sources of deployment challenges. Flexibility and scale are usually at opposite ends of a trade-off but both are desirable traits.
No special requirements, just an open mind.
I work at a small company (Idea Device) that has given me the opportunity to spend a large portion of the past five years seeing how small and large companies manage their operational processes and helping improve them. Before that I spent time at the Akshara Foundation realising that traditional software engineering is out of place with poorly defined goals and a varied user base. Along the way I was part of the Real Time Systems group at C-DAC where we built HPC clusters and used them.