Recently, it is noted that most of current Internet traffic comes from massive content delivery and the protocols used in it operate in ISP unfriendly way. In addition to them, narrowed position of ISP at Internet market requires a change from its traditional business model for ISP. Because massive content traffic over Internet seems inevitable, we suggest that active participation into content delivery service is a recommended direction for ISP instead of blocking or throttling content traffic.
From this point of view, this thesis presents an architecture for hierarchical DHT based network-level content delivery service provided by ISP. Hierarchical DHT provides an efficient and scalable data service architecture over hierarchical network topology. However, we point out that simple implementation of the architecture pertains several issues, such as longer query length from unbalanced identifier space between content and network node, overloading at a few nodes from the imbalance in content popularity and number of stored contents, and the ignorance of network status in routing decision. The last two issues are resulted from deterministic routing path selection in DHT. In order to resolve the aforementioned problems, we introduce heuristic based merging algorithm and dynamic cost based cooperative routing that exploits both of service load at each node and network cost of path from requester to candidate service node. In particular, we adopt caching and bloomfilter based cache-tracker mechanism in order to implement cooperative routing among nodes.
According to simulation based evaluation results, our proposed approach achieves around 40% shorter latency in both delivery and query compared to both of when flat DHT is applied and when caching is not applied. Further, it can alleviate overloading and content delivery through contested path efficiently.