Cloud storage systems can provide pervasive data access service to customers at anytime from anywhere, with very high data reliability and durability. Despite a large number of cloud storage services readily available so far, commercial cloud storage services are only used by a small fraction of personal users and enterprise users. We believe that a successful cloud storage service needs to have a strong economic incentive to attract active customers. The major challenge to the service provider is to recover its costs from charges to cloud users, and the charges should be maintained at a competitive level. This project aims at designing a new architecture for next-generation cloud storage services based on the rich theoretical foundation of economics, game theory, auction theory, and network coding. Different from existing cloud storage service, we design incentive mechanisms through online double auctions to attract cloud users to contribute their local resources such as disk storage space and outbound network bandwidth. Combining the resources at data centers and cloud users, we further design pricing models for the free retail market of storage service, such that cloud service providers can make profit, cloud users are charged fair prices, and overall market efficiency is maximized. In order to guarantee high data availability and durability with minimum overall cost, we propose to design a new network coding scheme as the underlying technology for reliable storage management. Different from existing coding schemes, our new network coding scheme is tailored for the hybrid storage structure composed of highly reliable storage devices at data centers and unreliable storage devices at cloud users. Considering the computational complexity of coding and decoding process, we propose to implement the network coding function by Graphics Processing Units (GPUs) which can improve the level of user satisfaction and significantly save the power cost at data centers.
|Effective start/end date||1/01/13 → 31/12/14|
Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.