Development and testing of database software for a more efficient handling of data in AI training

Important results from the project

Yes, the project achieved its goals. We developed a standalone, high-performance database based on our custom fork of OpenSearch and tested it with RISE. The results show up to 37% shorter response time, up to 57% more queries per second, and up to 35% lower energy consumption. The project also provided valuable insights into AI companies’ needs and confirmed the relevance of the problem in the industry.

Expected long term effects

The project is expected to reduce costs and energy consumption for companies handling large datasets, particularly in AI. By enabling faster database searches with fewer servers, both operational costs and climate impact are reduced. The results strengthen Sweden’s position in energy-efficient data technology and provide Vesiro with a platform for global growth and innovation

Approach and implementation

The project was carried out over one year, focusing on developing a standalone database based on Vesiro’s own algorithms, which we achieved by incorporating our algorithms in our own fork of OpenSearch. After development, the software was delivered to RISE, which tested it in the ICE Datacenter. The tests measured performance and energy consumption compared to OpenSearch. In parallel, a market study with AI companies was conducted to validate needs and create a basis for commercialization.