Open Source Summit + Automotive Linux Summit Japan 2019

There are many good reasons for storing your valuable data on premise, but also many good reasons for using the power of serverless cloud computing. What if we didn’t have to chose one against the other?

And even better, what of we could change our mind whenever we wanted?

In KubeCon2018 Huamin Chen & Yehuda Sadeh-Weinraub hooked Ceph into knative/kubernetes to do event driven image classification. But let us say we wanted the images in the “cat” bucket to be processed in the local kubernetes cluster while images in the “dogs” bucket to be processed by a shiny cloud function? And with minimal changes and in a future proof way.
In this talk Yuval Lifshitz would like to show how the recently introduced “publish-subscribe” mechanism in Ceph, when hooked into a message bus (e.g. rabbitmq), or directly to a serverless function in the cloud, could achieve the above.