Portal Digital Trends predicted that by 2020 the number of smartphone owners will reach 6.1 billion. And business data analytics platform Statista predicts that figure will rise to 7.5 billion by 2026. Many of the devices have modern processors, so they can be regarded as computing platforms.
Scientists in their work, published in the journal Communications in Computer and Information Science, said that, they say, smartphones can be used as computing nodes. In large numbers, of course.
In turn, computing nodes are a set of computers that have their own RAM and are controlled by their own operating system.
Ilya Kurochkin, co-author of the work, Associate Professor of the Department of Engineering Cybernetics at NUST MISIS, spoke in more detail about the study:
“Our work is dedicated to the use of mobile devices as providers of computing resources. The main idea of the project is to use the idle computing resources of mobile devices for various calculations, without prejudice to their direct purpose. This requires a series of different computational experiments to study the behavior of a grid system consisting of heterogeneous devices with different processor architectures and different operating systems, with subsequent tuning of various project parameters.”
One way or another, scientists managed to create a prototype of a mobile grid system that worked effectively with different types of personal devices. Testing was carried out on the BOINC platform. As its computing nodes can be not only mobile phones, but also tablets, personal computers and laptops. The computing application is available for several operating systems, including Android.
“In order to adjust the project parameters, we conducted a number of computational experiments on a test mobile distributed system. The found configuration made it possible to reduce the time of computational experiments, as well as increase the percentage of device loading and reduce the percentage of overdue tasks.
Ilya Kurochkin comments. According to scientists, the results obtained can be used for many months of experiments, for example, for training neural networks.
bbabo.Net