The system, designed to operate across a range of storage environments from corporate servers to medical imaging archives, uses intelligent algorithms to allocate data dynamically depending on user preferences and performance requirements.

It is already protected by a patent granted by the Polish Patent Office.

“Different types of media are needed depending on the application,” said Dr. Janusz Miroforidis of the Systems Research Institute PAS. “For example, slower drives can be used for storage, backup copies, where hundreds of terabytes of data are needed. However, a doctor working with medical imaging data would probably like to have quick access to the images. In this case, a different type of media is needed.”

Memory farms, typically composed of solid-state drives (SSD), magnetic disks, and even tape memory, are crucial for enterprises, financial institutions, design firms, healthcare facilities, and technology startups. These systems store, retrieve, and delete massive volumes of data in response to constantly shifting operational demands.

“In our solution, we take into account several criteria describing the work of a media farm,” Miroforidis continued. “For example, we can assume that users should have very quick access to their files, but we must remember that this will be associated with higher costs. In this situation, the preferences of the person managing the farm decide how it will work.”

Once the system is configured with input from the client—including the types and quantities of storage media and preferences like data access speed or energy savings—the software autonomously manages the operation of the memory farm in line with those priorities.

“During use, our system dynamically learns the characteristics of user requests to a given memory farm and, taking into account the programmed preferences of the administrator, stores files on specific carriers,” Miroforidis said.

The solution relies on core artificial intelligence technologies, including data clustering algorithms and multi-criteria optimization. It also utilizes high-performance computing methods executed on graphics processing units (GPUs), allowing the system to deliver real-time recommendations for memory allocation.

According to the research team, the system is scalable to farms of any size and is relatively inexpensive to implement. It was developed with support from the Polish National Centre for Research and Development using European Union funding.

The idea for the project originated with Infoklinika SA, a technology company that approached the Systems Research Institute PAS with a request to develop an intelligent controller for managing memory farm operations.

“We have to translate a given need or problem into formulas and mathematical models,” said Miroforidis. “We have to consider what methods of system analysis, mathematical modelling we can use. We translate what was expressed in natural language into the language of mathematics, computer science, optimisation. And this is how our product or prototype of a solution is created.”

PAP - Science in Poland, Ewelina Krajczyńska-Wujec

ekr/ bar/