LEVERAGING AI FOR ENHANCED APPLICATION SERVICE MONITORING

Abstract

When troubleshooting performance difficulties with distributed service-based cloud systems, it is essential to monitor their components and comprehend how they interact with one another. On the other hand, a lot of the current cloud monitoring solutions either don't work fully or need complex application and/or platform instrumentation, which makes them impossible to install on demand. Monitoring usually comes with a hefty price tag. This study presents the architecture of a cloud-based intelligent monitoring service network that can identify specific locations depending on predefined scenarios. Through a cloud server, our solution combines many CNNs that have been trained. These convolutional neural networks (CNNs) can recognize various features from webcam broadcasts. Figure 1 shows the three main components of our system: cameras, a cloud server for CNNs, and mobile terminals. In order to get the frames from the distant scene in real time, we employ a single webcam. We transmit these frames to our server in the cloud. In order to train CNN models, our cloud servers gather datasets from users. The findings of the recognition monitor are sent to the users' mobile terminals via the cloud server. Users may save a lot of computer resources by sharing trained models for comparable monitor applications. A hybrid CNN cloud server runs our whole monitoring system. In addition, we can optimize the number of virtual machines (VMs), physical locations of virtual machines (PMs), and servers (among other things) using artificial bee colony optimization. In comparison to current tools, our suggested model has much lower overhead, which results in a reduction of the monitoring-induced effect on reaction time of up to 89% and a reduction of resource consumption, including CPU and memory utilization, of up to 75%. An additional benefit is that the suggested deep learning module outperforms rival models in service identification by as much as 11%.