Traditional methods to classifying video across hot and cold storage devices – i.e. categorizing content based on access frequency or performance requirements – are being scrutinized for failing to adapt to dynamic shifts in content-aware factors. The bottom line is that streaming platforms may be able to improve delivered video quality by changing how they store different bitrate versions of content, rather than ploughing ahead with the current model of buying more storage or expanding CDN capacity. A new research study funded by the South Korean government and published in Cluster Computing explores how a new proposed scheme based on deep reinforcement learning (DRL) can manage hot storage zones—aka always-on servers—more effectively. DRL is a form AI in which a…