Summer 23 edition of Berkeley Datahub Newsletter
Get updates about the cutting edge work happening with Berkeley Datahub and across Berkeley Data Science Teaching Stack
Datahub Product Updates:
We are thrilled to bring you the latest updates and improvements to the Datahub architecture and its collaborative tools. In this newsletter, we will focus on the major improvements made to the infrastructure and teaching stack over the last 7 months.
Google Filestore Transition: As part of a series of efforts to improve the performance and stability of the instructional hubs, we transitioned from a self-managed Network File System (NFS) to Google Filestore. The majority of the downtime experienced in late 2022 was due to our NFS server crashing. The impact of this was that for all hubs, no user home directories were able to be mounted. This transition brings improved performance, scalability, and reliability for 15+ instructional hubs deployed at Berkeley.
Real-Time Collaboration (RTC): Real-Time Collaboration (RTC) refers to the ability for multiple users to collaborate and work together on a single notebook in real time. RTC is currently in the pilot stage and needs additional enhancements before it can be deemed a stable feature. We are seeking more volunteers to test the RTC functionality in a live classroom setting and provide feedback to the upstream Jupyter community. This input will help us enhance the RTC's functionality and reliability.
JupyterHub/bCourses Groups Functionality: Administrators can now use bcourses group affiliation to allocate resources through the Jupyterhub admin interface. This allows admins to effortlessly manage and allocate resources to specific groups/projects with resources as per their unique needs. For example, admins can allocate different RAM allocations based on the bcourses group membership. Read more about the functionality here.
Improvements to Calendar-based Scheduler: Improved the performance of the calendar scaler by refactoring the existing code by migrating to a new calendar library, as well additional error checking and logging. Previously, some instructors ran into issues with the calendar scaler as it didn’t provision expected resources automatically during the scheduled time. If you are interested to understand what is happening under the hood, you can read the technical documentation for the scaler.
Announcements/Updates:
JupyterLab 4.0: Number of accessibility improvements have been made in JupyterLab 4.0, including improved screen reader and keyboard navigation, more ARIA roles and labels, and a hamburger menu that collapses if there is not enough space to display all items. The stable version of Lab 4.0 was released in late May. You can read more about the features of Lab 4 in this blogpost here. We deployed the latest lab 4.0 release to a JL4 hub for Data100, which is one of the cutting edge educational Jupyterhub deployments. Students had a smooth experience during the past two weeks. If you are interested in deploying Lab 4.0 in your course setting then please do reach out to the Datahub team.
Jupyter Notebook 7.0: Reusable components from Lab 4.0 were used to build notebook 7.0 which will be the default notebook interface replacing classic notebook and retro notebook in the long run. Accessibility improvements from Lab 4.0 will get added to notebook 7.0 in the near term. Major courses including the EdX courses for Data Science will start using Notebook 7.0 during FA 23.
JupyterLite: JupyterLite was a lightweight version of the Jupyter platform designed to run in web browsers without the need for a traditional Jupyter server. It aimed to provide a streamlined Jupyter experience, particularly for environments with limited resources. JupyterLite allows users to interact with Jupyter notebooks and execute code directly within their web browser, making it easy to share and collaborate on notebooks without the need to install and run a full Jupyter server. It was primarily built using modern web technologies like WebAssembly and JavaScript. Data 8 notebooks can now be launched using JupyterLite and they can even get integrated to bcourses.
We hope these updates will enhance your Datahub experience. Stay tuned for more exciting developments in the coming months. Thank you for being a valued member of our Datahub community. If you have any questions or feedback, please don't hesitate to reach out to us via ds-infrastructure@lists.berkeley.edu.