Research On Enhanced Cloud Storage Service Based On Distributed Content Delivery

Personal cloud storage service,such as BaiduCloud,DBank,Dropbox,and Google Drive,are increasingly prevalent within the Internet community.It is estimated that subscriptions of personal cloud storage is projected to hit 1.3 billion in 2017.This tremendous number of users,along with overwhelming Internet traffic,brings the new challenge to the existing cloud storage system.As an attempt to alleviate the pressure of cloud datacenter and,at the same time,upgrade the data retrieving rate for users,we utilize distributed content delivery technique(P2P Network and Content Delivery Network)to propose a new hybrid system of cloud storage service.There are three main contributions in this project: 1)Coming up with a hybrid architecture for cloud storage service,in which CDN servers can redirect users' requests to either cloud datacenter or distributed nodes to retrieve the data.2)Based on the network model we build up,a workload allocation algorithm is proposed.The allocation algorithm takes the bandwidth saved rate and time saved rate as the principle to balance the workload between the datacenter and distributed nodes.Meanwhile,datacenter assigns its bandwidth to each workload according to the service pattern it belongs with.3)Since P2 P networks are made up of voluntary peers,the stability of service can hardly be guaranteed.We use CDN cache servers to fix this problem and design a Service-Oriented caching algorithm for P2 P networks.By contrast with former caching algorithms which only take the popularity of contents into account,the proposed caching algorithm prioritizes the contents that can not be well served by neighbor peers.To evaluate the proposed framework,plenty of simulations and evaluations are conducted.According to trace-driven simulations,the proposed system can achieve higher performance compared with several previous related works.The results show that the datacenter can save about 60% of the bandwidth on some occasion compared with the traditional system,and the improvement of data retrieving rate is more than 50% in average.In the future we plan to deploy the proposed system in a commercial cloud service provider,KylinCloud,to conduct more real experiments.The final goal of this project is to develop a new pragmatic architecture for cloud storage service and then benefit more network applications.