Memory-access-aware data migration techniques for low power embedded system

With the arrival of the Internet of Things (IoT) era, the emergence of new applications to improve various aspects of daily life is encouraged. Most Internet of things devices are small-scale, and battery power sources have improved the mobility of these devices. In this way, execution at low power is an important issue because it is necessary to extend the battery life. In order to improve the performance of small-scale embedded systems, we propose a data migration method for transferring read-dominant data from SRAM to Flash memory. We trace memory accesses, analyze memory access patterns, and separate read-dominant data from the read/write data. Then, the read-dominant data is relocated to the Flash memory sector. These procedures are able to reduce the energy, power, and current consumption for accessing the data in SRAM. Experiments showed that the proposed methodology achieves reduction of power and current consumption compared with conventional storage, which keeps all data in SRAM. Data migration technique could manage efficiently energy and power in IoT device.