Marc Verhoeyen

Video content recommendation: An overview and discussion on technologies and business models

This paper presents several aspects, from technologies to business models, of content recommendation for video related solutions that range from IP television (IPTV), including linear programming television (LPTV) and video-on-demand (VoD), to online video. Video content recommendation is becoming increasingly important because of the continuously increasing amount of video content available to end users. After considering some end user requirements, an analysis is provided of the most important content recommendation technologies as described in literature and implemented by many start-ups. The paper also deals with evaluation criteria for a content recommender system related to user expectations, support of different scenarios (e.g., new content, new users), and marketing and business requirements. We describe the overall architecture into which the content recommendation functionality fits, as well as its interfaces with other network components, external databases, social networks, and applications. Finally, we discuss a dedicated network provider play, the business opportunities, and several business models for content recommendation.

Optimizing for video storage networking with recommender systems

Driven mainly by its adoption as a new media distribution platform for content providers and its ubiquitous availability for the end users media production and consumption, the Internet is rapidly reshaping. In particular, the stakeholders in the content distribution market are considering exploiting content delivery networks (CDNs) to play a key enabling role allowing them to become part of related value chains. In this paper we discuss how such CDNs rely on autonomous algorithms to optimally use the storage resources, i.e., reducing bandwidth on feeder links, while providing quality of experience (QoE) to the end user. A model is presented to allow the comparison of algorithms that rely on measuring content popularity versus new recommender based algorithms that base caching decisions on predictions of individual end user behavior. The performance and the dynamics of these algorithmic components are assessed based on a theoretical demand model for a cache deployed at one of the various levels in a tree-based access network, inclusive of a single user cache. This scientific analysis allows to carefully pitch some considerations towards infrastructure network providers that deploy such storage networking.