This article introduces a method for realizing the Gaussian activation function of radial-basis (RBF) neural networks with their hardware implementation on field-programmable gaits area (FPGAs). The results of modeling of the Gaussian function on FPGA chips of different families have been presented. RBF neural networks of various topologies have been synthesized and investigated. The hardware component implemented by this algorithm is an RBF neural network with four neurons of the latent layer and one neuron with a sigmoid activation function on an FPGA using 16-bit numbers with a fixed point, which took 1193 logic matrix gate (LUTs—LookUpTable). Each hidden layer neuron of the RBF network is designed on an FPGA as a separate computing unit. The speed as a total delay of the combination scheme of the block RBF network was 101.579 ns. The implementation of the Gaussian activation functions of the hidden layer of the RBF network occupies 106 LUTs, and the speed of the Gaussian activation functions is 29.33 ns. The absolute error is ± 0.005. The Spartan 3 family of chips for modeling has been used to get these results. Modeling on chips of other series has been also introduced in the article. RBF neural networks of various topologies have been synthesized and investigated. Hardware implementation of RBF neural networks with such speed allows them to be used in real-time control systems for high-speed objects.
For effective management of cloud data center resources there is a need to develop and implement simple and robust methods that take into account the data center power consumption, user Service Level Agreement compliance, and different conditions of resource utilization. In this paper, the authors analyze the possibility of application of the reinforcement learning method to cloud data center resource management. Due to the intensive changes of the workloads and different conditions of resource utilization the data center resource management problem should be solved dynamically in an online manner. To address such problem, the authors propose the method of dynamic virtual machine placement based on the reinforcement learning. The proposed method takes into account the power consumption and the number of SLA violations while producing control impacts. Besides, the method considers the changes in the resource utilizations to make a decision on whether to switch underloaded physical servers to the sleep mode in order to reduce the power consumption. The proposed reinforcement learning agent is based on the Q-learning approach which allows to determine the optimal policy for managing the physical servers without creating an environment model and preliminary information about the workload.
Cloud data centers are most dynamic systems in a modern digital world. To deliver the high-performance and fault-tolerant IT services to end users effectively it is necessary to develop new methods for data center resource management while adapting to the emergence of new requirements. In this paper, the authors refine and evaluate the previously proposed method for cloud data center resource management based on the reinforcement learning approach. The proposed method takes into account the power consumption and the number of SLA violations in the management policy. The power consumption is managed by switching physical servers to active or sleep state depending on current utilization of three resources: CPU, memory, and network bandwidth. The proposed reinforcement learning agent allows to determine the optimal policy for managing the physical servers without creating an environment model and preliminary information about the workload. The evaluation results show that the proposed method allows to decrease the SLA violation time, to serve more VM schedule requests when the number of VMs is changing frequently, and to decrease the utilization of data center network due to decreased number of migrations.
A widespread use of the cloud computing paradigm has increased the necessity and significance of improving the management efficiency of cloud data centers. Special attention is paid to solving cloud resource management problems. Due to the intensive changes of virtual machine (VM) workloads and different conditions of resource utilization the VM placement and migration problems should be solved and optimized continuously in an online manner. To address such problems the authors present an approach to continuous new VM allocation and VM migration. The authors also evaluate a particular policy of the VM allocation in a data center using an adaptive genetic algorithm. The proposed Adaptive Software Defined approach to the cloud data centers management is implemented by using the policy selector, that allows to select different algorithms or policies for resources and virtual machines management in order to adapt to the impact of disturbing influences.