Big data integration could involve a large number of sources with unpredictable redundancy information between them. The approach of building a central warehousing to integrate big data from all sources then becomes infeasible because of so large number of sources and continuous updates happening. A practical approach is to apply online query scheduling that inquires data from sources at runtime upon receiving a query. In this paper, we address the Time-Cost Minimization Problem for online query scheduling, and tackle the challenges of source permutation and statistics estimation to minimize the time cost of retrieving answers for the real-time receiving query. We propose the online scheduling strategy that enables the improvement of statistics, the construction of source permutation and the execution of query working in parallel. Experimental results show high efficiency and scalability of our scheduling strategy.