HYBRID METHOD FOR HORIZONTAL AND VERTICAL COMPUTATIONAL RESOURCE SCALING

Abstract

B a c k g r o u n d . Increasing the efficiency of cluster computing resources while maintaining the established QoS levels is a critical task in IT infrastructure management. Dynamic management of computing resources, in particular vertical and horizontal scaling, are tools that allow automating the processes of adapting applications to dynamic loads. The aim of the work is to improve the efficiency of existing scaling methods by combining them. M e t h o d s . A hybrid scaling method using the coordination module is proposed. This module coordinates the operation of vertical and horizontal scaling components based on the given constraints, priorities, and the current state of the system. The coordination module aims to increase efficiency of the both components and prevents inconsistency in the combination of the number of instances and resource requests for each instance. R e s u l t s . To achieve the given objective, a hybrid method of vertical and horizontal scaling using priority-based component coordination was developed. Priority configuration affects the order of components operation. In the case of horizontal-vertical order, the non-prior vertical component does not affect configuration of the prior horizontal component. The developed method is evaluated based on modeling the operation of an application with a load containing a constant seasonality. The experiments demonstrate a 65% reduction in the unprofitable reservation of cluster computing resources compared to static requests. C o n c l u s i o n s . The developed method can be used to increase the efficiency of resource utilization in clusters under dynamic loads compared to basic scaling methods. In further research, it is necessary to evaluate the method using real infrastructure. It is also necessary to investigate the work of a hybrid method using a predictive approach