Table 4 Software used in the validation experiment

Log data collection workflow controller (also called workflow-C)

To take control of log data collection workflow and instruct other controllers (described below and in section V-2) to enable autonomous data collection

Openstack

Virtualization infrastructure management

KVM

Hypervisor of the cloud

Resource controller (scratch)

To control the resource amount allocated to the VM according to the workflow-C′ instruction

Environmental condition controller (Stress-ng [25])

To simulate the environmental conditions of the host according to the workflow-C′ instruction

Training image dataset generation tool (scratch)

To generate a training data set of a given number of images with a given number of pixels and color channels according to the workflow-C′ instruction

Neural network generation tool (based on Tensorflow [26]))

To generate image recognition models of given layers, given number of neurons in each layer and activation functions according to the workflow-C′ instruction

Neural network training tool (based on Tensorflow [26])

To train the generated model with the generated image dataset using the given amount of resource on the VM

Training time monitoring tool (based on Tensorflow [26])

To record the execution time of for the generated model and training data set using the given amount of resource on the VM