Carbon Footprint and Carbon-Aware Selection of Document-Oriented, Relational, and Hybrid Object-Relational Storage for Image-Based Time-Series Workloads in Green IoT
This paper benchmarks the carbon footprint of MongoDB, PostgreSQL/TimescaleDB, and a PostgreSQL+MinIO hybrid for image-based time-series workloads in green IoT, revealing a resolution-dependent emission crossover where MongoDB is most sustainable for lower resolutions (up to 1440p) while the hybrid approach excels at higher resolutions (4K and above) due to decoupled storage, ultimately providing a carbon-aware decision framework for selecting storage architectures in sustainable building monitoring.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
In the modern world, our buildings are becoming increasingly smart. Cameras and sensors watch over offices, factories, and homes, capturing a continuous stream of images and data to monitor safety, energy use, and activity. This data does not just sit idle; it must be stored, organized, and retrieved instantly, often for years. To do this, facilities rely on digital storage systems, the software libraries where this information lives. For decades, engineers chose these systems based on speed and cost. However, as the world shifts toward sustainability, a new question has emerged: how much energy does it take to keep this data alive, and what is the carbon cost of that energy? Every time a server writes an image to a disk or reads it back, it consumes electricity. That electricity, depending on where the server is located, generates carbon emissions. The choice of storage software is no longer just a technical decision; it is an environmental one.
A team of researchers set out to measure exactly how different storage systems handle the growing demand for high-resolution images. They focused on three common approaches used in the industry. The first is a document-oriented system, which stores each image and its details together in a single, flexible package. The second is a traditional relational system, which organizes data into rows and columns, often keeping large images tucked away in a separate part of the same database. The third is a hybrid approach, which keeps the small details in the main database but sends the heavy image files to a specialized storage service designed just for large objects. The researchers wanted to know which of these methods burns the least energy and produces the least carbon as image quality increases from standard definition to ultra-high definition.
To find the answer, the researchers built a controlled experiment using a single computer running three different storage setups, one after the other. They used a tool that measures the actual energy drawn by the computer's processor and memory, converting that energy use directly into carbon emissions based on the local power grid. They tested the systems with images ranging from small, low-resolution pictures to massive, six-thousand-pixel-wide photographs, simulating the kind of data a smart building might generate. They measured how fast each system could store the images, how quickly it could retrieve them, and, most importantly, how much carbon was emitted during every step of the process.
The results revealed a surprising twist: there is no single "greenest" system for every situation. Instead, the best choice depends entirely on the size of the images being stored. For smaller, lower-resolution images, the document-oriented system was by far the most efficient. It stored these files with such little overhead that it produced a fraction of the carbon emissions of the other systems. In fact, for the smallest images tested, it was nearly twenty times cleaner than the most polluting alternative. This happened because the system could compress these small files so effectively that they took up almost no space and required very little energy to write.
However, as the image resolution grew, the story flipped. When the images reached high-definition sizes, the document-oriented system began to struggle. It tried to keep the massive image files inside its main storage package, which caused the system to work much harder and consume significantly more energy. At the highest resolutions tested, this system actually failed completely, unable to store the images at all because the combined size of the data exceeded a hard limit built into its design. In contrast, the hybrid system, which separates the image files from the main database, became the clear winner for high-resolution work. By offloading the heavy image files to a dedicated storage service, it avoided the energy penalties that plagued the other systems. At the highest resolutions, the hybrid approach produced roughly half the carbon emissions of the traditional relational system.
The traditional relational system, which kept everything in one place but used a special technique to move large files out of the main data path, acted as a middle ground. It was never the absolute best performer, but it was never the worst either. It handled the full range of image sizes without failing, making it a reliable choice for situations where the image resolution might vary or where simplicity is more important than squeezing out every last drop of efficiency. The researchers also looked at the long-term environmental cost of the storage space itself. Because the document-oriented system expanded the size of the data so much at high resolutions, it required significantly more physical storage hardware. This means that over time, the manufacturing and disposal of the extra hard drives needed for that system would add a substantial, hidden carbon cost to the project.
The study concludes that facility managers and engineers cannot simply pick one system and stick with it forever. If a building relies mostly on standard-definition cameras, the document-oriented system is the most sustainable choice. But if the infrastructure uses high-resolution cameras for detailed inspection or surveillance, the hybrid approach is the only way to keep emissions low. The researchers provided a clear decision guide: choose the document system for small images, the hybrid system for large ones, and the traditional system if the mix is uncertain. This work transforms the abstract concept of "green computing" into a practical rulebook, showing that the most sustainable technology is the one that matches the specific nature of the data it is asked to hold.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.