چكيده به لاتين
Power consumption is one of the major issues in cloud computing systems which may lead to a serious increase in electricity consumption and cause global warming. Virtual machine technology used in cloud computing systems, has a lower utilization than container technology under unpredictable workloads, therefore container technology has been drawing more attention in cloud computing recently. Due to the major popularity of containers and microservice orchestration, the need to evaluate power consumption and performance trade-offs is vital more than ever. Since measurement-based evaluations require too much budget and time, analytical modeling evaluations are in favors.
Lack of related works and analytical models for evaluating power consumption and performance trade-offs in microservices orchestration is one of the main motivations of this research. Reducing datacenters costs, environmental pollution and waste of natural resources are other relevant motivations of this research.
In this thesis, an analytical model using stochastic activity networks is proposed to evaluate power consumption and performance in microservice orchestration. The proposed model can evaluate power consumption and performance trade-offs as well as the effects of dynamic scaling in microservice orchestration on power consumption and performance which has been ignored in most of the current related works. The proposed model includes different parts of cloud systems including containers, virtual machines and servers. Dynamic scaling and reactive scaling in microservice orchestration are considered in the model as well. After experimenting different scenarios, the steady-state results produced by Mobius are presented.
