ILM and tiered storage system are designed to address the challenge of achieving balance between cost and storage performance. However, both of them are hard to implement fully automatic data migration by traditional solutions for which are mainly relying on administrators experience and need huge manual work for data migration according to storage configuration and IO access patterns. This paper proposes a novel bi-directional migration policy FDTM based on block-level data valuation and fully automatic migration process. FDTM aims to get a trade-off between storage QoS and migration costs by introducing double thresholds to narrow the migration scope of block-level data objects. Experiment and analysis show that FDTM is efficient at block-level data migration comparing with traditional migration policies. In addition, it could help pave the way to implement tiered storage system with fully automatic data migration.