The Study Of Locality-aware And Interest-aware Aroma-CDN Model And Its Key Problems

Dedicated Content distribution networks (CDN) are expensive technologies for file-sharing and distributing the content of web-servers to large audiences while distributing the internet traffic more or less evenly disposed. Unfortunately, non-profit websites (e.g., related to education organizations, social organizations, scientific associations, etc.) often can not afford the expenvive expenses of deploying and administrating a dedicated CDN infrastructure. Thus, their under-provisioned servers easily become overloaded with large mount of queries. What these websites need is a distributed content distribution infrastructure/solution that can deliver the content at large scale without the large costs of traditional CDNs. P2P-CDN which combines P2P with CDN technologies is an appropriate solution to meet the requirements above.For the purpose of effective use of network infrastructure, every node participated in P2P-CDN solution can perceive itself' s network locality, find its required content object nearly, and get the content object whithout the knowledge of the underlayer network topology. In order to get one-hop delivery service, the basic idea is to let each peer be connected to a content overlay which represents a cluster of peers that have the same interest and reside close to each other. Peers keep the content they retrieve and later serve it to other peers that are close to them in same locality. When a new client requests some content from a website, a locality-aware multiple DSHT infrastructure quickly finds a content overlay that has the content available. All peers in a given locality that maintain the content of a particular website build an unstructured content overlay. Within this overlay, peers are allowed to gossip each other at a reasonable traffic to maintain accurate information despite churn.In this paper, after several P2P CDN key technologies have been analysed, we propose a hybrid CDN model supporting locality-awareness and interest-awareness, called Aroma/Earoma-CDN. We focus on three tasks listed below.1. As to probe the Round Trip Time (RTT) between nodes and also from node to the Landmark servers, we present a RTT/locality probing solution named Large Variance-Adaptive Probing with Predictable Refreshing Cycle (LVAP/PRC), whereby nodes partition themselves into bins such that nodes that fall within a given bin are relatively close to one another in terms of network latency.2. We propose a two-layered P2P-CDN architecture consisting of a Distributed Sloppy Hash Table (DSHT) structured overlay and unstructured petal overlays to provide hybrid locality-aware query routing with limited background traffic cost. We can get fast lookuping speed and good scalability while avoiding hot-spot congestion and tree saturation by the sloppy DHT. Within a petal, peers use gossip protocols with Logistic function adaptive cycle to exchange information about their content and contacts rapidly, avoiding file repeatly copy across-ISP networks.3. We propose EAroma-CDN, which dynamically adapts to increasing numbers of participants in order to avoid overload both directory peers and petals. We describe how to to maintain EAroma-CDN under churn in face of dynamic changes and failures, by relying on adaptive gossip protocols and alocality-aware maintenance protocol for our novel DSHT. Simulation-based evaluation shows that Aroma/EAroma-CDN yields a high performance gains under large population with high churn.