Research On Data Deployment For Cloud Data Center Based On Erasure Code

Cloud computing has been attracting the attention of industry and academia. As an important infrastructure of cloud computing, data centers consists of large quantity of interconnected servers, which store large amount of data to provide service to end users.Three concerns exist in deploying data in the cloud data centers. First, latency is an important metric to measure the service quality, which is also important for the development of cloud computing. Second, the cloud service providers want to reduce the storage cost, while traditional multiple data copies strategy is making the data centers have to provide more and more storage capacity. Third, it is urgent for the cloud service provides to reduce the energy cost.The main contributions of this thesis are as follows:First, we propose a data deployment approach LSCD (Latency and Storage Cost driven data Deployment) to the problem of striking the tradeoff of latency and storage cost in a data center. All the datas are stored by erasure code and a full copy is provided for each data piece that is frequently accessed so that the access latency can be reduced. Based on Minimum Cost Maximum Capacity, LSCD finds the appropriate server to accommodate each data piece and erasure code data chunk.Second, we propose an algorithm LEDD (Latency-aware and Energy-efficient Data Deployment) to achieving the balance of latency and energy cost for the data placement in multiple data centers. Each data piece not only is stored using erasure code, but also has a full copy. Based on Particle Swarm Optimization, LEDD iteratively provides the accommodation of each data copy and erasure code data chunk.