In a Databricks workspace, where does MLflow store experiment metadata (parameters, metrics) by default?
Select an answer to reveal the explanation.
Short Explanation and Infographic
Databricks manages all the plumbing for you — experiment metadata goes into a managed backend store, so you never have to configure a database server the way you would with open-source MLflow.
Full explanation below image
Full Explanation
In Databricks, MLflow is a managed service with a hosted tracking server. Experiment metadata (runs, parameters, metrics, tags) is stored in Databricks' internal managed backend — backed by Delta tables in the workspace's internal metastore. This is fully managed, highly available, and requires zero configuration. Artifact files (models, plots, large files) are stored in the workspace's default artifact root, typically cloud object storage (S3, ADLS, GCS) managed by Databricks. In contrast, open-source MLflow requires users to configure their own backend store (SQLite, PostgreSQL, MySQL) and artifact store (S3, Azure Blob, GCS). The local filesystem answer would mean data is lost when the cluster terminates. /mlflow/experiments/ in DBFS is not where managed metadata is stored.