In a cloud data center network, virtual machines (VMs) can be dynamically created and terminated, causing the fact that fragmented resources cannot be further utilized. To solve the problem, the server consolidation is proposed to minimize the number of active servers. Although the approach can reduce resource fragmentation at particular time, it may be over aggressive at the price of too frequent VM migration. Therefore, a novel provident resource defragmentation algorithm is proposed to reduce the unnecessary VM migration and maxmize the profit by re-consolidating fragments into available continuous resources. An optimization problem is formulated, and a heuristic method for obtaining near-optimal results is developed. Extensive numerical results confirm that the framework provides the highest profit and significantly reduces the VM migration cost.