Optimizing Color-Based Cooperative Caching in Telco-CDNs by Using Real Datasets

Abstract

Content Delivery Networks (CDNs) play a vital role in efficient content distribution inside the Internet to reduce network traffic and improve users’ experience. Because CDNs are usually located outside Internet Service Providers (ISPs), they cannot decrease traffic inside ISPs and on peering links between ISPs and CDNs. Hence, there is a considerable need to deploy CDNs inside ISPs to tackle this issue. Additionally, these CDNs are called Telco-CDNs since we are in the control of ISPs. Traditional caching policies widely used in CDNs can be applied to Telco-CDNs. Nonetheless, the policies are inefficient in the context of Telco-CDNs in that network operators of CDNs have no knowledge of underlying network infrastructure while those of Telco-CDNs do. The fact leads to the emergence of caching algorithms in Telco-CDNs. Color-based caching strategy together with its routing algorithm is regarded as the most effective one with acceptable computation overhead. In principle, contents will be assigned to color tags and periodically re-colorized every interval in the approach. Since the previous study used a simulated dataset following gamma distribution, the characteristics of this dataset did not change every interval. For that reason, there was no experiment to verify the efficiency of the algorithm when the characteristics of the dataset varied. In this paper, we conduct numerous experiments to look the aspect over. The experimental results show that not only the color-based approach still remains prominently effective in comparison with Least Frequently Used (LFU) every interval, but also the strategy with periodical re-colorization outperforms the one without re-colorization. Moreover, the prior research only took account of users’ interests in a global manner rather than geographically local regions, which is difficult to attain the optimizing traffic reduction. Thus, we also propose an iteration of the color-based caching strategy by making use of the insight of users’ preference based on regional areas to optimize traffic. The experimental findings reveal that traffic can considerably be reduced for all local areas, especially by up to 27.3%. To sum up, the proposed extension of the color-based caching strategy surpasses the traditional color-based one in practice.