Pending-interest-driven cache orchestration through network function virtualization

Emerging information-centric networking architectures drives a wider focus of content caching mechanisms. Most recent efforts have been attempting for server and user friendliness, i.e., reduction in content-server load and content-delivery time. Those solutions require the prior knowledge of cache state and content popularity exchanged among routers. Challenging tasks are pushed to routers since their explicit coordination is required. Instead of burdening routers, but still keeping the effectiveness of in-network caching, we move the explicit coordination task from routers to a cache orchestrator where the offline optimal caching policy is computed while leaving the simplest cache decision to the routers. By means of the network function virtualization, this paper proposes a cache orchestration lifecycle including three parts, i.e., name hit caching (NHC) policy, pending time history and network cache orchestration. By exploiting NHC policy, a router simply caches the content whose name matches its assigned content name set. Evaluation through simulations demonstrates that NHC policy achieves the lowest content-server load and content downloading time and the highest cache hit ratios in all routers comparing with leaving copy everywhere, probability caching and content-popularity-based cache orchestration policies.