SeoulTech Scientists Develop Ultra-Lightweight Memory Manager That Transforms Embedded System Performance

The revolutionary software could dramatically boost the performance of billions of smart devices

SEOUL, South Korea, Sept. 2, 2025 /PRNewswire/ -- Embedded systems such as Internet of Things (IoT) devices and single-board computers possess limited memory and processing power, necessitating the effective management of these constraints. This makes Linux—a flexible and cost-effective platform—promising for embedded systems. Indeed, Linux-based operating systems, including Ubuntu Core, Raspberry Pi OS, BalenaOS, and OpenWrt, are commonly used for a wide variety of embedded devices. However, ptmalloc—its default memory allocator—is often unable to satisfy the needs of all applications. While experts have proposed alternatives such as jemalloc, tcmalloc, and mimalloc for improved memory management, these general-purpose dynamic memory allocators suffer from heavy memory consumption, vast library sizes, complexity, and eventual performance degradation. This highlights the urgent requirement for new lightweight options.

Addressing this issue to reduce complexity and optimize performance, researchers led by Dr. Hwajung Kim, an Assistant Professor of Smart ICT Convergence Engineering at Seoul National University of Science and Technology (SeoulTech), Republic of Korea, have developed LWMalloc, a lightweight and high-performance dynamic memory allocator specifically designed for resource-constrained environments. Their novel findings were made available online on 12 February 2025 and have been published in Volume 12, Issue 12 of the IEEE Internet of Things Journal on 15 June 2025.

LWMalloc is based on a lightweight data structure and possesses a deferred coalescing policy and dedicated small chunk pools for memory allocation optimization. Its data structure helps reduce metadata overhead for compact and efficient implementation. Notably, the DC policy postpones redundant operations until allocation, thus lowering the execution overhead and maintaining efficiency as well as low-response times. Furthermore, dedicated small chunk pools segregate small memory requests common in dynamic allocation patterns into fixed-size pools, facilitating their O(1)—constant time complexity—allocation.

The researchers demonstrated the superiority of LWMalloc through extensive experimental real-world applications on Raspberry Pi Zero W, Raspberry Pi 4, and Jetson Orin Nano. "Our proposal outperforms ptmalloc, the default allocator in Linux, achieving up to 53% faster execution time and 23% lower memory usage. With a compact implementation of only 530 lines of code and a 20-KB size, significantly smaller than ptmalloc's 4838 lines and 116 KB, LWMalloc achieves an effective balance between performance and memory efficiency, making it highly suitable for resource-constrained environments," remarks Dr. Kim.

LWMalloc can benefit any embedded or IoT system that operates under strict memory and performance constraints. These include consumer electronics, such as smart TVs, set-top boxes, home appliances, mobile and wearable devices, automotive systems with real-time constraints, as well as edge computing nodes handling AI or data processing workloads.

In the long term, efficient memory allocators like LWMalloc can extend device lifespans, reduce energy consumption, and enable more complex applications to run on low-power hardware. According to Dr. Kim: "This could make high-performance features accessible on affordable consumer devices, reduce e-waste, and improve the responsiveness and reliability of everyday embedded systems. Moreover, as IoT and edge computing continue to expand, such lightweight allocators will be critical in ensuring the scalability and sustainability of connected devices worldwide."

Reference
Title of original paper: LWMalloc: A Lightweight Dynamic Memory Allocator for Resource-Constrained Environments
Journal: IEEE Internet of Things Journal
DOI: 10.1109/JIOT.2025.3541247

About the institute Seoul National University of Science and Technology (SEOULTECH)
Website: https://en.seoultech.ac.kr/

Contact:
Eunhee Lim
82-2 - 970 - 9166
[email protected]

SOURCE Seoul National University of Science and Technology (SEOULTECH)