Demonstration of ElasticNotebook: Migrating Live Computational Notebook States

Computational notebooks (e.g., Jupyter, Google Colab) are widely used for interactive data science and machine learning. However, existing notebook systems lack the functionality of reliably and efficiently persisting thenotebook session state consisting of user-defined variables (e.g., processed datasets, ML models), hence the termination of a session often leads to loss of work. In this demo, we introduce a new notebook system, ElasticNotebook, that offers live migration of session states via computational checkpointing/restoration for notebook systems (e.g., Jupyter Notebook, Colab). ElasticNotebook's frontend allows users to configure the periodic creation of session state checkpoints, which can then be restored at will through a drop-down menu. ElasticNotebook's backend utilizes novel lightweight monitoring techniques to find a reliable and efficient way (i.e., replication plan ) for replicating session states when requested. This demo will showcase ElasticNotebook's ability to preserve the user's work progress in Jupyter Servers by replicating their session state in two common use cases: live migration across machines and resumption after termination.