Since our last blog post on January 2018, we have changed a lot the Datalayer architecture. Back in 2018, we had chosen for Apache Zeppelin for its good integration with Big Data frameworks like Apache Spark and competely replaced the existing Angular.js user interface with a home-brewed React.js implementation to integrate with the Kubernetes Control Plane. While rolling out more and more features on top of our former version 0.0.1, we have been intrigued in February 2018 by JupyterLab being announced to be ready for users. Back in time, in July 2016, JupyterLab was positioned as the next generation of the Jupyter Notebook.
We knew that Jupyter had a strong academic user base, so we tought it was worth to question our stack choice. Looking deeper, we have discovered that JupyterLab was built with Typescript and that JupyterHub, the multi-user flavor of Jupyter, was running on Kubernetes. Instead of building our own as we had done for version 0.0.1, we decided to learn, connect and use existing Jupyter solutions. We have kept the underlying Kubernetes cluster, but everything on top of that has changed. We have thrown away all our work and started from scratch, using the well-know Jupyter ecosystem.
Apart from the technical "re-platforming", we have connected to a lot of business and academic users. Going out of the office and talking with users has brought up a lot of ideas and needs. Our main focus is now to scope those needs and to release a MVP with our today building blocks and technical architecture.