Ephemeral Runtime Storage and Local File Persistence Limits

Local Disk Volatility in Hosted Runtimes

Google Colab utilizes ephemeral virtual machines for notebook execution. While the environment provides a local file system located at /content/, this storage is volatile. Any data written to the local disk is permanently deleted when the runtime is recycled, whether due to a manual "Disconnect and delete runtime" action or an automatic session timeout.

To maintain data across different sessions, users typically mount Google Drive. However, there is a functional gap when dealing with large datasets or temporary files that exceed the available local disk space but do not require long-term archival in cloud storage.

Resource Constraints

• Local storage is cleared upon runtime termination.
• Session durations are subject to automatic disconnection based on inactivity.
• Resource allocation varies by subscription tier.

What are the specific triggers that force a runtime recycle over a simple restart? Is there a documented method to extend the lifecycle of the local /content/ directory without mounting an external drive?