Recently, data service companies analyzing IoT data to provide valuable knowledge emerge as a promising business model. To sustain such business model, the establishment of a data service platform for IoT data sharing and data service delivery becomes a necessity. Today, blockchain-based platforms are believed to be reliable but have poor system performance, considerable storage overheads and high energy consumption. MEC technology is then applied to help IoT devices participate in the blockchain network. However, the considerable loading of blockchain operations turns MEC hosts into a performance bottleneck, instead. Therefore, this paper exploits the hierarchical blockchain architecture in the proposed IoT data service platform, called HB-IoDS, to significantly reduce the storage overheads and service providing time. Furthermore, to alleviate the heavy burdens on MEC hosts, HB-IoDS provides the adaptive Proof-of-Work (APoW) consensus algorithm that derives an appropriate difficulty to achieve PoW condition according to device capability. By this mean, resource-constrained IoT devices can also perform blockchain operations which are usually regarded as energy-consuming. Finally, simulation results show that HB-IoDS outperforms existing blockchain-based platforms in terms of system performance, energy consumption, and storage costs. As a result, the proposed HB-IoDS is the most practical IoT data service platform.