Multi-Leveled Hierarchical Control to Optimize Workload of a Service-Oriented Platform

nowadays, one of the directions of development of cloud systems is a creation of open platforms, providing resources for deploying third-party services. Users of such resources may deploy their own services, implemented with a wide range of technologies, receiving guaranteed performance and availability. To provide high availability and optimal utilization of hardware resources such platforms are built with a complex hierarchical infrastructure. As complexity of such systems increased, due to inability of control methods on different levels of hierarchy becoming ineffective, an issue with performance appeared. A complex approach to control is needed. This work is focused on researching the probability of making a hierarchical control system that allows choosing optimal methods and influence different system's elements on different levels, depending on current state of the system. A prediction of a probable workload of the system is also taken into consideration. The testing platform allowed conducting simulation experiments to determine threshold values that will influence system's state and to evaluate cost of control methods at certain levels of hierarchy. The received results show that it is highly probable to use the multi-leveled hierarchical control to achieve optimal performance level.