Two-criteria technique for the resource-saving computing in the fog and edge network tiers

Introduction. At present, the concepts of fog and edge computing are used in a wide range of applications of various kinds. One of the key problems in the organization of computing in groups of mobile devices that make up the edge/fog layer is the mission assurance based on battery power availability. In this context, a lot of developments aimed at energy saving of device systems have been presented to date. However, one important aspect remains beyond the consideration of the problem of resource saving, namely, the issue of saving the residual resource of a computing device. The aim of this research is to formalize the workload distribution problem as two-criteria optimization problem, and to develop the basic solution technique.Materials and Methods. Within the framework of this article, an approach to resource saving is proposed. It is based on the evaluation of two device criteria: battery life and residual resource of a computing device. The residual resource of a computing device can be estimated using the probability of failure-free operation of the device, or as the reciprocal of the failure rate, taking into account that the exponential law of failure distribution is used in the simulation. From this, a model of the problem of two-criteria optimization is formulated, taking into account the dynamics of the network topology in the process of performing a user mission. The topology dynamics is reflected in the model as a sequence of topologies, each of which corresponds to a certain period of time of the system operation.Results. Based on the proposed model of the two-criteria optimization problem, a method was proposed for resource saving in the edge and foggy layers of the network. It reflected the specifics of the dynamic layers of the network, and also took into account the importance of the criteria for estimating the consumption of device resources. An experiment was conducted to evaluate the impact of the method of distributing tasks over a network cluster on the probability of failure-free operation of devices and on the average residual resource.Discussion and Conclusions. The conducted experiment has demonstrated the feasibility of using the developed method, since the distribution of tasks among executing devices had a significant impact (up to 25 % according to the results of the experiment) on the average residual resource of a computing device.