Russian Scientists Use Software Optimization to Reduce Power Consumption

Scientists from Volgograd State Technical University (VolgSTU) have developed a method to optimize software that will help reduce the power consumption of computer systems by four, the university believes.

In today's world, high-powered computing systems are used to perform precise calculations, , in particular server arrays. Volgograd State Technical University (VolgSTU) experts explained that this makes most devices’ operations particularly power-consuming.

Significant resources are required not only to ensure uninterrupted operation of computing clusters but also to cool them due to overheating. Scientists noted that the task of reducing power consumption and the rate of wear-and-tear on the devices has been set in almost all major payment centers in the country.

While this problem of energy intensity previously belonged to hardware developers and the design of more cost-effective equipment, more recently, the search for new power solutions has involved software developers, in particular specialists from VolgSTU.

The researchers have developed a method to optimize the software's cache memory, that is, a buffer with fast access to the most relevant information it contains. They then applied the technology to software for complex mathematical transformations.

“We calculate the probability of a cache miss, and based on the results, we attempt to improve the caching algorithms. This scheme can be applied to any energy-intensive software until the program is written,” Alla Kravets, professor at the Computer-Aided Design and Exploratory Construction Department of VolgSTU Faculty of Electronics and Computer Engineering, said.

She explained that the first step of the method involves dividing the set of memory access operations into non-overlapping subsets. Then specialists get subsets of operations with equal values of cache miss probability. As a result, the total number of operations is determined, and a formula is generated that estimates the program’s running time. The results of their study were published in the Energies journal.

“The resulting power consumption has been reduced many times over, and it can be reduced by up to four times. In fact, optimized software versions consume more energy. However, due to the significant reduction of the program execution time, power consumption is also reduced,” Vitaly Egunov, associate professor at the VolgSTU Department of Electronic Computing Machines and Systems, added.

The next task of the research team is to further develop the proposed method, as well as to generalize the proposed approach to other subsystems of computing devices, the scientist continued.

This research is part of a VolgSTU strategic project within the framework of the "Priority 2030" program.

The university development program includes four strategic projects, including the "Center for Digital Scientific and Educational Projects and Developments" and "Technologies for Industrial Innovation Cluster."