KJCKorea
Journal Central

In traditional flash memory-based storage devices, the processing speed of the flash memory chips exceeded the transfer speed of the host interface, limiting the storage device's performance to the I/O bandwidth of the interface. However, the situation fundamentally changed as storage device interface technology evolved from SATA and SAS to NVMe. The NVMe interface leverages the PCIe bus to provide tens of times more transfer bandwidth than its predecessors, significantly exceeding the processing speed of flash memory. Consequently, the performance bottleneck has shifted from the interface layer to the flash memory processing layer, demanding a new approach to storage device design. This paper proposes an architecture to address this issue. It organizes multiple flash memory chips operating independently within an SSD into logical groups and applies an internal RAID technique to each group, enabling parallel processing of high-bandwidth data. The proposed method not only maximizes system throughput by effectively utilizing the increased bandwidth but also provides robust recovery capabilities against bit errors, chip failures, and data loss during large-scale data processing through RAID's parity-based error correction. Consequently, this research presents an efficient storage system solution that simultaneously satisfies high performance and high reliability in next-generation high-speed interface environments.

With the global increase in electricity demand and growing concerns over climate change, improving energy efficiency and reducing carbon emissions in buildings have become critical challenges. Effective energy management during the building operation phase requires real-time, data-driven analysis and control. This paper proposes an AIoT (Artificial Intelligence of Things)–based intelligent energy digital twin platform to address these needs. The proposed system adopts a three-layer architecture consisting of a sensor layer, an edge layer, and a data management and service layer, enabling real-time acquisition and processing of environmental and occupancy data. A YOLO-based object detection model is deployed on edge devices to estimate occupancy from camera streams, which is integrated with environmental sensor data to accurately reflect space-level energy usage characteristics. In addition, a digital twin–based energy model is developed to replicate building energy consumption patterns in a virtual environment and derive rule-based energy-saving guidelines. The proposed platform provides a scalable framework that can be extended to renewable energy and external energy systems, contributing to building-level energy optimization and future smart campus and smart city applications.

This study aimed to examine the effects of reflective writing activities utilizing an AI writing platform on learners' writing self-efficacy, writing performance, and self-reflection attitudes. Employing a mixed-method approach, the research statistically verified pre- and post-test changes in the three variables among students enrolled in "Writing and Communication" at A University. Additionally, learners’ writing experiences were explored in depth through the analysis of reflection journals. Participants initially drafted responses to 10 writing prompts focused on self-understanding and subsequently engaged in iterative revisions incorporating AI-generated feedback. The results indicated statistically significant improvements in all three variables following the intervention. Specifically, AI feedback enabled learners to perceive writing as an achievable task, while criteria-based feedback enhanced writing quality and performance. Furthermore, iterative revisions focused on themes of self-understanding promoted personal reflection and strengthened learners’ self-reflective attitudes. This study is significant in that it provides empirical evidence to inform the integration of AI into future university writing curricula.