Recently on a project I had to make changes to a underlying portion of the sites architecture to move sessions in Zend Framework from file storage to database storage. However this affected a piece of the architecture. Node.js, which manages all our real time interaction, looked at sessions at the file level. This was quite a easy transition for the function as it was abstracted away in a function call so the theory was to just replace the function "guts" with a new component.
The post shows the code he came from (which pulled in the PHP session file and extracted the session data manually) over to a new database-based version that selects from the SESSIONS table and pulls out the data. It's based on the table having an "id" column and the Zend_Auth namespace it uses.