This work presents a Dynatrace OneAgent extension for gathering NVIDIA GPU metrics using NVIDIA Management Library (NVML). The extension integrates GPU metrics into an industry-leading platform for Application Performance Management extending its capability of monitoring important business workloads to the GPU-oriented computational nodes. A practical approach for acquiring and processing NVML metrics via Python bindings is described. The work also proposes and discusses implementation of helper applications for convenient simulation of performance problems in a multi-tier web application. These applications are then used in combination with OneAgent-based monitoring and appropriate configuration of Dynatrace platform for web application monitoring. Next, an end-to-end production-like scenarios are presented, which exemplify extension usefulness in test setup resembling a real world implementation. The extension has been released on GitHub under MIT license.
NVIDIA GPU Performance Monitoring using an Extension for Dynatrace OneAgent
