Abstract
We consider the task of content replication in distributed content delivery systems used by Video-on-Demand (VoD) services with large content catalogs. The prior work in this area focuses on the setting where each request is generated independent of all past requests. Motivated by the fact that most popular VoD services offer recommendations to users based on their viewing history, in a departure from existing studies, we study the setting with time-correlation in requests coming from each user. We use a Markovian process to model each user’s request process. In addition to introducing time-correlation in user requests, our model is consistent with empirically observed properties of the request process for VoD services with recommendation engines. In the setting where the underlying Markov Chain is unknown and has to be learned from the very requests the system is trying to serve, we show that separating the task of estimating content popularity and using the estimates to design a static content replication policy is strictly sub-optimal. To prove this, we show that an adaptive policy, which jointly performs the task of estimation and content replication, outperforms all policies that separate the task of estimation and content replication.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have