In some components you can specify file URL attribute as a reference to some resource on the file system. Also external metadata, lookup or DB connection definition is specified as reference to some file on the filesystem. With CloverETL Server there are more ways how to specify this relation.
All relative paths in your graphs are considered as relative paths to the root of the same sandbox which contains job file (ETL graph or Jobflow).
Sandbox URL allows user to reference the resource from different sandboxes with standalone CloverETL Server or the cluster. In cluster environment, CloverETL Server transparently manages remote streaming if the resource is accessible only on some specific cluster node.
See Using a Sandbox Resource as a Component Data Source for details about the sandbox URLs.