Research On Data Management And Maintaince In Social Networks Based On Incremental Partitioning

The emergence and development of computer network brought us a new way to communicate, fast and convenient. Social network, as the latest form of communication online, is one of the most popular services, attracting billions of people. With so many users, managing their information and interactions seems to be a challenge.In this thesis, after deeply investigating on the characteristics of data from social networks, we propose an interactive directed graph to describe user interactions in social network. We compare our model with some already exist ones and argue how it can be used to describe user actions in social networks.Based on the interactive directed graph, a model to manage social network data is proposed. In order to accelerate accesses to "hot" records and delay persistency of newly registered users, the model uses two sub-structures, one for persistent data and the other for cached ones. The model uses some parameters to quantify the cost of migration and replication, and also load of each partition.On top of our model, a strategy to manage and maintain data from social networks is introduced. The strategy is based on an incremental partitioning method, which handles each of the common user requests from social network applications in an effective way. Besides the partitioning method, the strategy also deals with some periodic maintenance and reacts to some special requests and emergencies.To evaluate our effort, we use some data from a real social network website. The experiment shows that our method can achieve a high data locality with a low cost. After analysing the result and cost, we do some discussion on the parameters and their effects.