Cost-aware and Time-Varying Request Rate-based Dynamic Cache Provisioning over CCDN

Today’s Internet users depend on Cloud-based Content Delivery Network (CCDN) providers to efficiently use resources and minimize content delivery latency. However, time-varying request rates make incoming requests unpredictable, resulting in higher miss rates in CCDNs. Thus, dynamically allocating cache instances to mitigate request rate spikes is necessary to maintain Quality of Service (QoS). Therefore, by utilizing the Request Count Window (RCW) cache and a dynamic instantiation mechanism, this article aims to efficiently manage a changing request window and store objects according to a caching policy. To improve cache performance, we propose a new dynamic selective caching policy that caches only the most frequently requested objects over the past L requests. To further enhance dynamic cache provisioning and reduce the duration of resource allocation, a variable time slot mechanism based on the current request rate is introduced. Finally, we present a cost model that accounts for caching and delivery costs due to missed occurrences. The proposed method is extensively compared with the Least Recently Used (LRU) caching policy and the RCW cache with selective caching based on K^th request cache insertion policies. Experimental results demonstrate that the proposed method outperforms the LRU, selective caching on the values 1 and 2 of K by 22.04%, 12.11%, and 9.31%, and by 29.50%, 16.71%, and 3.03% in terms of the average total cost using permanent cache sizes of 2 and 5, respectively.