CDN 환경에서의 동적 콘텐츠 캐싱을 위한 콘텐츠 뷰카운트 예측 및 인기 콘텐츠 판별

Abstract

The content delivery network (CDN) provides an environment in which a large amount of contents can be transmitted stably and quickly from suppliers to consumers. However, as high quality and large capacity contents are produced and consumed, a more efficient content caching technique is required in CDN. It is a common phenomenon that a small number of contents occupy the most demands in CDN. To relieve the possible problems caused by this phenomenon, we propose a hybrid content-level viewcount forecasting method based on moving average and Prophet for dynamic contents caching to support an optimized dynamic contents caching in CDN. Based on the predicted viewcount, each content is classified into either hot or cold contents for each day to help dynamic content allocation: hot contents are allocated in a fast but expensive storage while cold contents are allocated in a slow but inexpensive storage. Experimental results show that the proposed method yields better hot/cold contents classification performance than benchmark methods.