Viaccess-Orca is field-testing an AI model that can scan the next two weeks of a programme guide, predict the CPU loads needed for the video processing across the shows, and provision sufficient cloud capacity to cope with peak workloads, then provide the operations team with a guide to what they can expect or automatically provision the capacity needed.
The AI model is designed to ensure service providers or content owners do not over-provision cloud capacity (wasting money and causing unnecessary carbon footprint) and do not under-provision, either (causing performance and UX issues). The model was trained by scanning historical programme guide data and the expected benefits are now being verified with real playout.
The characteristics of shows are analyzed (e.g. their semantics) and this understanding of the content is cross-referenced against CPU loads for video processing and against viewing figures. With this learning, the model can look at the future-facing EPG, consider the characteristics of the listed programmes, and make predictions about audiences and CPU loads.
Regular patterns are noted, and changes to regular patterns are also flagged. Cloud capacity requirements can be right-sized to a projection that will be more accurate and more dynamic than if the operations team predicted their cloud workloads without the AI. Viaccess-Orca was demonstrating the AI model last month at IBC.
