kunpeng-perf-monitor
kunpeng-perf-monitor is a Prometheus Exporter for Kunpeng servers. It is deployed in DaemonSet mode on each node in a Kubernetes cluster to collect node-level hardware and kernel performance metrics. The metrics collected by kunpeng-perf-monitor can be obtained directly through the exposed /metrics endpoint. You can also connect kunpeng-perf-monitor to Prometheus through configuration to implement visualized detection of related metrics.
DevKit Collector is a kunpeng-perf-monitor's collection component implemented based on the Kunpeng DevKit Tuner CLI. This component periodically collects the CPU pipeline, cache, and memory access data, and converts the collection results into Prometheus metrics for analyzing performance issues such as CPU pipeline bottlenecks, memory access efficiency, and bandwidth.
Figure 1 kunpeng-perf-monitor architecture
Constraints:
Only Kunpeng 950 processors are supported.
Parent topic: Cloud Native Infrastructure