CONSERVATION OF FOG LAYER BROKER COMPUTE NODE RESOURCE IN DATA PROCESSING APPLICATIONS

Abstract

In accordance with the concept of fog computing, when processing large amounts of data, the computing workload shifts from the “cloud” to the edge of the network, to the nodes of the fog layer. At the same time, however, a very small number of published works deal with the issues of reducing with that the residual resource of peripheral devices, which, as a rule, are not of a high computational power comparing with devices in data centers. The reliability indicator, the probability of failure-free operation are also associated with such a characteristic as the average residual resource – the value of the reciprocal of the failure rate, and in general, the decreasing trend of the value characterizes the safety of the computing resource of the device and the time during which its operation will be expedient due to the allowable failure rate. Modern implementations of the concept of fog computing suggest the presence of a so-called cloud-fog broker, which is entrusted with the functions of computing scheduling that can be performed by nearby fog layer nodes. Obviously, in this case, the cloud-fog broker operates with an increased load, and it is quite reasonable to raise the question of saving the resource of the node. The article is concerned with the study of comparing the “selfish” strategy of distributing the computational load over the nodes of the foggy layer in the network and the approach based on the choice of a node for computing based on the simulation of the average residual resource. The proposed algorithm makes it possible, depending on the preferences in the safety of the resource of the node, to implement the choice of device. The simulation performed demonstrates the expediency of including the fog layer broker in the subset of considered candidates for load distribution even in the case when, in accordance with the “selfish” approach, it is more profitable for the broker to transfer calculations to the next node. The latter is also relevant for the case when it is a priority to reduce resource consumption for a group of devices, including a broker