vNV-Heap: An Ownership-Based Virtually Non-Volatile Heap for Embedded Systems
Markus Elias Gerber, Luis Gerhorst, Ishwar Mudraje, Kai Vogelgesang, Thorsten Herfet, Peter Wägemann · 2025
Read vNV-Heap: An Ownership-Based Virtually Non-Volatile Heap for Embedded Systems by Markus Elias Gerber, Luis Gerhorst, Ishwar Mudraje, Kai Vogelgesang, Thorsten Herfet, Peter Wägemann (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2501-17707.txt.
Introduction
Markus Elias Gerber, Luis Gerhorst, Ishwar Mudraje, Kai Vogelgesang, Thorsten Herfet, and Peter Wägemann build a memory heap for computers that run on harvested energy. Their 2025 paper serves the batteryless internet of things: devices that sip power from light or vibration, die often, and must resume exactly where they stopped. Saving state costs energy, so only changed data should be written to non-volatile memory, which demands precise change tracking.
Their heap, vNV-Heap, borrows the ownership discipline of the Rust language, a compile-time system that knows exactly which code may modify each object. That knowledge virtualizes persistence: objects act volatile while the heap quietly guarantees they survive power cuts. Static analysis bounds every operation's worst-case energy cost. Tests on an ESP32-C3 board showed the abstraction is energy-efficient.
Catalog record
| CREATOR | Markus Elias Gerber, Luis Gerhorst, Ishwar Mudraje, Kai Vogelgesang, Thorsten Herfet, Peter Wägemann |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 56 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2501-17707
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
vNV-Heap: An Ownership-Based Virtually Non-Volatile Heap for Embedded Systems Markus Elias Gerber, Luis Gerhorst, Ishwar Mudraje, Kai Vogelgesang, Thorsten Herfet, Peter Wägemann Abstract . The Internet of Batteryless Things might revolutionize our understanding of connected devices by harvesting required operational energy from the environment. These systems come with the system-software challenge that the intermittently powered IoT devices have to checkpoint their state in non-volatile memory to later resume with this state when sufficient energy is available. The scarce energy resources demand that only modified data is persisted before a power failure, which requires precise modification tracking. We present vNV-Heap , the first ownership-based v irtually N on- V olatile Heap for intermittently powered systems with guaranteed power-failure resilience. The heap exploits ownership systems, a zero-cost (i.e., compile-time) abstraction for example implemented by Rust, to track modifications and virtualize object persistence. To achieve power-failure resilience, our heap is designed and implemented to guarantee bounded operations by static program code analysis: For example, the heap allows for determining a worst-case energy consumption for the operation of persisting modified and currently volatile objects. The evaluation of our open-source implementation on an embedded hardware platform (i.e., ESP32-C3) shows that using our heap abstraction is more energy efficient than existing approaches while also providing runtime guarantees by static worst-case bounds. DOI: 10.1145/3735452.3735534 ISBN: 979-8-4007-1921-9/25/06 Conference: Proceedings of the 26th ACM SIGPLAN/SIGBED International Conference on Languages, Compilers, and Tools for Embedded Systems; June 16–17, 2025; Seoul, Republic of Korea Proceedings of the 26th ACM SIGPLAN/SIGBED International Conference on Languages, Compilers, and Tools for Embedded Systems (LCTES ’25), June 16–17, 2025, Seoul, Republic of Korea CCS: Computer systems organization Embedded and cyber-physical systems 0 0 footnotetext: ©2025 Copyright held by the owner/author(s). This is the author’s version of the work. It is posted here for your personal use. Not for redistribution. The definitive version was published in LCTES ’25. 1. Introduction Intermittent computing has the potential to enable sustainable IoT ( IoT ) devices that are privacy-friendly and low-maintenance ( Ahmed et al., 2024 ) . To date, most IoT systems rely on batteries to power themselves. However, this is not sustainable as batteries have a high carbon footprint ( Dehghani-Sanij et al., 2019 ; Peters et al., 2017 ) and must be frequently replaced because their capacity wears out over time ( Han et al., 2019 ; Yang et al., 2018 ) . Having to replace batteries not only becomes an infeasible maintenance effort in the presence of many small devices but is also simply not possible for certain remote applications ( Attarha and Förster, 2024 ) . Intermittent computing solves these issues by harvesting energy from the environment (e.g., from solar cells) and buffering it in an energy storage device (e.g., capacitors). However, this approach introduces challenges for application and system developers, as the power may fail unexpectedly at any time. Still, the device has to recover from a power failure quickly to be able to make reasonable forward progress. While early approaches to intermittent computing periodically create checkpoints, other approaches use reactive or just-in-time checkpointing ( Maeng et al., 2017 ; Maeng and Lucia, 2018 ; Jayakumar et al., 2014 ; Balsamo et al., 2015 ; Balsamo et al., 2016 ) . In these systems, the hardware notifies the OS ( OS ) when energy is about to run out. Asynchronous interrupts, which occur prior to an imminent power failure, allow the OS to only create checkpoints when actually required. This reactive checkpointing speeds up execution when energy is available but, in turn, requires the OS to limit the amount of volatile state (i.e., RAM ( RAM ), processor registers, and device state) that has to be persisted during a checkpoint. Furthermore, embedded applications (including on intermittent systems) have to date been severely constrained in their memory usage ( Alonso et al., 2015 ; Association et al., 2004 ) , complicating development and limiting their features. This not only affects novel applications such as confidential AI ( AI ) ( Müller et al., 2024 ; Warden and Situnayake, 2019 ; Brown, 2022 ; Terven et al., 2023 ; Dattu et al., 2025 ) , but also more traditional tasks like data compression and queues/buffers in the networking stack ( Vogelgesang et al., 2025 ; Schoeberl and Pedersen, 2018 ) . To solve this problem of constrained memory, NVM ( NVM ) has the potential to increase the amount of available memory, therefore enabling more feature-rich embedded applications. This is because it does not require refresh cycles and is cheaper ( Köhler et al., 2024 ) . However, to date, it cannot be efficiently used on embedded systems as simply replacing volatile RAM with NVM comes with severe performance overheads ( Jayakumar et al., 2014 ) and with increased energy demands to access data ( Maioli and Mottola, 2021 ) . This is the case even with the most high-end NVM technologies ( Rabenstein et al., 2023 ) that have yet to trickle down from servers to embedded systems. To make efficient use of NVM , the OS has to transparently use fast volatile RAM as a buffer to cache frequently accessed data from the larger NVM (referred to as swapping in Linux). Unfortunately, the traditional approach to implementing this with virtual memory requires a MMU ( MMU ), which is not available on the low-power processors required for intermittent computing. To summarize, embedded and intermittent systems require appropriately-grained modification tracking and swapping (i.e., virtual memory ) implemented without relying on MMU . The former is required to limit the amount of volatile data that has to be persisted when a power failure is imminent. The latter is required to increase the amount of (virtual) memory available in order to allow for more feature-rich applications. Existing hardware-agnostic solutions to this problem either track modified state with insufficient granularity ( Balsamo et al., 2015 ; Balsamo et al., 2016 ; Bhatti and Mottola, 2016 ) or are too fine-grained, therefore inducing unacceptable memory- and execution-time overheads ( Ahmed et al., 2019 ) . More flexible solutions require the programmer to manually insert unchecked acquire/release calls for memory regions into their programs to enable modification tracking ( Sliper et al., 2019 ) . To allow for predictable, safe, and adaptable modification tracking and swapping for intermittent systems, we propose to leverage modern programming-language features to allow for predictable resource behavior through a safe interface that is checked by the compiler. For this, we design an ownership-based vNV-Heap ( vNV-Heap ) to replace traditional virtual-memory systems with an alternative that addresses the domain-specific requirements of embedded systems. Using the programming-language concepts of borrowing and ownership, our vNV-Heap can virtualize resources, guarantee safe non-volatile memory usage (thereby extending memory safety from volatile RAM to NVM ), precisely track modifications, offload unused objects, and limit modified state. Our contributions are five-fold: (1) Ownership-based vNV-Heap : We propose the concept of ownership-based virtual memory to bring features traditionally only available with MMU to embedded systems. To the best of our knowledge, we are the first to propose the use of ownership-based resource safety in virtually non-volatile heaps. (2) Adaptability: We design the vNV-Heap to adapt to the application requirements. Programmers can