Issues in next generation data management

New trends in data sharing are emerging: data objects are becoming larger and larger, or the data must remain very small but associate with time constraints. Conventional database management strategies are not suitable for these new trends, in which there can be too much data to store or there can be time constraints for disseminating small data objects.;There can be a mix of requests in many on-demand broadcast environments: some of the requests with time constraints, others without. Existing data broadcast strategies typically only consider either how to minimize the wait time of the requests or how to minimize the number of deadlines missed. In this dissertation, we presented a cost model for mixed-type broadcast environments considering both response time and number of deadlines missed. Given the Markov Decision Process model of the system, we then proposed two scheduling strategies based on this cost model: Maximum Paid Cost First and Maximum Value Gained First. Further, we discussed the transaction missing rate in the broadcast environment. We proposed a new data server scheduling strategy for read-only transactions with deadlines after introducing a new cost model to measure the transaction system performance.;In a data intensive computing environment, shared data is typically replicated to improve the job response time and system reliability. In this dissertation, we studied two new metrics of data availability to evaluate the reliability of the system. Then, we proposed four new strategies for limited replica storage that maximize the data availability based on a file weight and prediction function. Meanwhile, the fairness of the scheduling is ignored when too much attention is paid to the system turnaround time. We proposed a new approach for data intensive computing that is designed to improve the system turnaround time in a fair manner. We introduced a Remote Data Access Element in data intensive computing that relieves the system from busy-waiting by adapting a new Data Backfill Scheduling strategy. Later, one novel Sliding Window Replication scheme was then presented. Lastly, we proposed a new performance metric which measures the balance between the scheduling fairness and system turnaround time.