Dynamic Predictors for Content Selection in Content Distribution Networks
Independent thesis Advanced level (professional degree), 20 credits / 30 HE creditsStudent thesis
Caching in Content Delivery Network is one of the leading methods for saving and providing Quality of Services to users in terms of low latency experienced when requesting multimedia resources. Caching allows a parsimonious use of bandwidth for service providers to have a scalable system and avoid network congestions. Most of the research has focused to save contents in CDN in order to meet the restriction of memory and bandwidth consumption relying on optimal content placement problem and cache policy. The most common policy used to cache content is based on the content's popularity, i.e., the request frequency. The availability of predictions in the requests of content would allow to optimally cache content. However, how to analyze past content requests to have consistent prediction of future data requests is an open and challenging problem. In this master thesis, this has been addressed by considering data mining, which is a multidisciplinary technique involving theoretical and practical data analysis. Dynamic predictors are designed and proposed to retrieve inherent content information for improving the prediction of the content item selection. Numerical results show that the proposed method achieves good results in term of hit ratio, i.e., low prediction error, which might be used by CDN designer and might be a potential input for the optimal content placement problem.
Place, publisher, year, edition, pages
2014. , 58 p.
Electrical Engineering, Electronic Engineering, Information Engineering
IdentifiersURN: urn:nbn:se:kth:diva-142844OAI: oai:DiVA.org:kth-142844DiVA: diva2:704604
Master of Science in Engineering - Electrical Engineering