Research And Implementation Of Personalized Content Scheduling Model In Edge Caching Network

The Internet industry is developing rapidly,and Internet traffic is also increasing.Compared with Internet services at the beginning of this century,today’s Internet applications require mass distribution of personalized content and require extremely low request latency.Traditional content delivery networks(CDNs)struggle to cache sparse personalized content and meet the latency requirements of these applications.In this regard,this paper establishes an edge cache model suitable for personalized content,which schedules personalized content in a targeted manner,that is,performs cache prefetching and eviction actions on it to achieve a higher cache hit rate.The main work of this paper includes:(1)For personalized video content distributed by Internet video applications,this paper proposes a video edge caching model suitable for personalized video representation.Internet video applications distribute video files with different video representations with different video encoding formats,resolutions,and frame rates for different users.This model innovatively integrates video conversion actions such as video encoding and conversion,video super resolution,video interpolation,and video compression,and responds to a large number of requests that require different video representations with a single or a small number of video files.With the help of the active cache grooming method at leisure,the model described in this paper also makes full use of the peak-valley nature of requests to complete content scheduling,that is,pre-download and preconversion of video files.The cache hit ratio of this model is significantly higher than that of mainstream edge cache models that only perform video super-resolution conversion.(2)For personalized content distributed by streaming,this paper proposes a general personalized edge caching model assisted by streaming recommendation system,which is suitable for various personalized content such as videos,images,and text.This model innovatively introduces the output of a streaming recommendation system to predict what the user will request.Based on time series forecasting technology,this model uses a user’s request history to personalize the time when they will make subsequent requests.Finally,the model sorts all content and actively performs content pre-download and cache eviction actions to schedule personalized content.Compared with the mainstream recommendation system-edge cache joint optimization model,this model can still achieve a high cache hit rate under sparse user interest.In addition,this streaming recommender helper method can also be cascaded with the video edge caching model described above to achieve higher performance video edge caching.(3)For crowdsourced edge cache networks,this paper proposes a personalized edge cache model based on user behavior prediction.This model establishes a two-layer structure of self-owned edge network and crowdsourced edge cache network,which assigns user temporary identifiers to users for crowdsourced edge cache nodes to associate different requests of the same user,while avoiding malicious nodes from tracking users across nodes.It also innovatively observes and exchanges lists of available content to more accurately predict sparse and timevarying user interests.Finally,the edge cache node observes the continuous online status of each user to predict their subsequent active status.This model does not rely on external information sources,protects user privacy,and achieves a higher cache hit ratio than the performance bounds of the benchmark model.