Video on Demand is the original streaming use case. A finished asset — a movie, an episode, a training clip, a marketing video — is encoded, packaged and stored once. When a viewer presses play, the player fetches the manifest, picks a rendition, and starts pulling segments from a CDN edge. Nothing is being created in real time, which means every operational variable is calmer: packager load is predictable, segments live forever in cache, you can re-encode mistakes overnight without losing viewers.
VOD divides commercially into SVOD (subscription, like Netflix), AVOD (ad-supported, like YouTube), TVOD (transactional pay-per-view), and FAST (Free Ad-supported linear channels, which are technically VOD assets stitched together to look live). Each model has different demands on personalization, ad insertion and DRM, but they share the same underlying packaging — a CMAF-based ABR ladder, multi-DRM, captions in WebVTT or IMSC1.
The hard problems on VOD are catalog size and the long tail. A big OTT service has hundreds of thousands of hours of content; only a few percent are watched in any given week. That means cache hit ratios are dominated by warming and eviction policy, not by raw bandwidth — and that picking the right ABR ladder for each asset (one renditiontoo many wastes storage, one too few hurts quality) is a per-title decision, increasingly automated with content-aware encoding.
