Search

Cloud computing, now a mature computing service, provides an efficient and economical solution for processing big data. As such, it attracts a lot of attention in academia and plays an important role in industrial applications. With the recent increase in the scale of cloud computing data centers, and improvements in user service quality requirements, the structure of the whole cloud system has become more complex, which has also made the resource scheduling management of these systems more challenging. Thus, the goal of this research was to resolve the conflict between cloud service providers (CSPs) who aim to minimize energy costs and those who seek to optimize service quality. Based on the excellent environmental awareness and online adaptive decision-making ability of deep reinforcement learning (DRL), we proposed an online resource scheduling framework based on the Deep Q-network (DQN) algorithm. The framework could make a trade-off of the two optimization objectives of energy consumption and task makespan by adjusting the proportion of the reward of different optimization objectives. Experimental results showed that this framework could effectively be used to make a trade-off of the two optimization objectives of energy consumption and task makespan, and exhibited obvious optimization effects compared with the baseline algorithm. Therefore, our proposed framework can dynamically adjust the optimization objective of the system according to the different requirements of the cloud system. [ABSTRACT FROM AUTHOR]

Cooling systems provide a safe thermal environment for the reliable operation of IT equipment in data centers (DCs) while generating significant energy consumption. Therefore, to achieve energy savings in cooling system control under dynamic thermal distribution in DCs, this paper proposes a multi-setpoint cooling control approach based on deep reinforcement learning (DRL). Firstly, a thermal model based on the XGBoost algorithm is constructed to precisely evaluate the thermal distribution in the rack room to guide real-time cooling control. Secondly, a multi-set point cooling control approach based on the deep Q-network algorithm (DQN-MSP) is designed to finely regulate the supply air temperature of each air conditioner by capturing the thermal fluctuations to ensure the dynamic balance of cooling supply and demand. Finally, we adopt the extended CloudSimPy simulation tool and the real workload trace of the PlanetLab system to evaluate the effectiveness and performance of the proposed approach. The simulation results show that the proposed control solution effectively reduces the cooling energy consumption by over 2.4% by raising the average air supply temperature of the air conditioner while satisfying the thermal constraints. [ABSTRACT FROM AUTHOR]

Cloud computing is a computing service provided on demand through the Internet, which can provide sufficient computing resources for applications such as big data and the Internet of Things. Before using cloud computing resources, users need to lease the corresponding virtual machines in the data centers, and then submit the tasks to the leased virtual machines for execution. In these two stages, how to choose the optimal data center for resource deployment and task submission is particularly important for cloud users. To tackle the problem that it is difficult to select optimal data center to lease and use virtual machines in distributed data centers, we proposed data center selection algorithms based on deep reinforcement learning. In the stage of virtual machine lease, aiming to minimizing user costs, we considered both the cost of computing resources and the cost of network communication. Then, we used a deep enforcement learning algorithm to obtain the shortest communication path between the users and the data centers,and solve the problem of network communication costs which are difficult to calculate. We achieved an optimal selection of data centers and effectively reduce the overall user cost. In the stage of virtual machine use, to improve quality of service for the users, we use a deep reinforcement learning algorithm to obtain an optimal task scheduling strategy. This effectively solved the problem that user tasks cannot be effectively scheduled due to dynamic changes in the type and size of user tasks and in the state of the virtual machines. Moreover, the proposed scheme reduces the overall task completion time. [ABSTRACT FROM AUTHOR]