Methods and systems for creating and retrieving scalable media files using a chunking model. A plurality of network abstraction layer (NAL) units representing a scalable media presentation are grouped into a plurality of video chunks wherein each chunk represents only the NAL units of one layer and a specific time period. The plurality of video chunks are then stored in one or more scalable media files along with indexing information which provides information for retrieving the NAL units from each video chunk. In this manner, a client can request only the video chunks (and thus NAL units) that it needs or wants to achieve a specific bandwidth.