PeCo: Minimizing Bandwidth Cost for CDN Systems

To provide high-quality content caching services for a large number of users, CDN service providers face enormous traffic pressure and immense bandwidth costs. This paper studies the traffic allocation between the CDN nodes in the offline situation to reduce network transmission pressure and bandwidth cost. We first propose an algorithm, PeCo-2, that can achieve an optimal solution in polynomial time in a simplified case of the problem, which only allocates traffic between two nodes. Then, based on the idea of node combination, we extend PeCo-2 to two heuristic algorithms PeCo-N and PeCo-Huffman in the general case. Then the simulation demonstrates that our algorithms outperform all previous methods in most cases on a practical data trace from a large commercial CDN system. And in most extreme cases, the consistency of experimental results can be guaranteed.