z/VM Neural Network Processor
Detailed metrics on the IBM Z Artificial Intelligence Accelerator
The dashboard shows various performance metrics of the AI Accelerator introduced with z16 Telum processor to perform various neural network tensor operations. For z/VM® this function can be exploited by a Linux on IBM Z workload.
Different charts on the dashboard show the average execution time and waiting time per NNPA operation, the utilization by AI Accelerator and the total utilization from the z/VM workload. The Linux on z/VM workload is shown as breakdown in AI Accelerator and General Purpose CPU usage to provide insight in the type of workload. An additional chart shows total CPU usage by Linux guest, making it easier to distinguish the AI workload in a configuration with other workload.
For z17 the charts also show the usage of AI Accelerator on other chips in the machine and provide insight in the performance improvements by avoiding waiting for a busy AI Accelerator.
A detailed description is provided as Dashboard for z/VM Neural Network Processor.
This dashboard is licensed by IBM under the Apache 2.0 License and is provided ‘as is’ without warranty, representation, support, maintenance or an obligation to issue updates.
Data source config
Collector config:
Upload an updated version of an exported dashboard.json file from Grafana
| Revision | Description | Created | |
|---|---|---|---|
| Download |