To make use of big data, various NOSQL data stores have been deployed, such as key-value stores and column-oriented stores. NOSQL data stores typically achieve a high degree of scalability, while specialized for some specific purposes; thus, Polyglot persistence that employs multiple NOSQL data stores complementally is a practical choice toward a high diversity of application demands. We assume various NOSQL data stores running on database servers are accessed by clients via a network. This paper aims to improve performance of the Polyglot persistence by introducing an FPGA-based 10GbE network interface (NIC) and In-Kernel NOSQL Cache (IKNC) implemented in the NIC device driver. IKNC stores query results as a key-value form in a host memory, and the requested data can be returned to clients immediately if the query result has been cached. In the IKNC keyvalue pair, the key is a hashed value of a given search query and the value is a query result of the search query. Existing works have focused only on key-value stores, while that for column-oriented stores that support range queries (e.g., scan operation) has not been addressed. In this paper, we also propose two cache strategies of IKNC for column-oriented stores. In our experiments, Apache HBase is running on an application layer, while our IKNC with the proposed cache strategies is implemented on an FPGA-based NIC and its device driver. A significant performance improvement is achieved by the proposed IKNC and pros and cons of the proposed two cache strategies are demonstrated.