With the advent of network functions visualization and software defined networking, cloud content distribution network (CDN) providers can auto-scale their virtual CDN appliances in order to meet changing demands for commercial and user generated content services in a cost and energy efficient manner. However, existing caching policies, constrained to work with dedicated CDN resources and designed to maximize local cache hit rates, do not exploit the elasticity of virtualized cloud environments to adaptively guarantee service requirements with minimum cost. In this paper, we design and evaluate MC3 (MinCostCloudCache), an adaptive distributed caching strategy whose fundamental goal is to guarantee content service requirements while minimizing the use and associated cost of the shared physical infrastructure. MC3 estimates the global benefit of caching an object at a network node using only locally available information. The caching benefit is flexible and adaptive to the particular content service requirements, and is aware of the behavior of neighbor network caches, creating effective cache cooperation using only local information. Through simulation, we show how MC3 not only reduces the experienced average delay with respect to existing caching policies, but it also uses significantly less storage and transport resources, leading to increased revenues and reduced operational costs.
展开▼