Docs

Application Performance Management Plus

Product overview

Application scenarios

APMPlus provides benefits such as low-cost integration, end-to-end monitoring, and comprehensive coverage for AI scenarios across the entire chain. This enables you to quickly locate issues and improve service quality. This topic describes the typical application scenarios of APMPlus.

AI cloud-native inference

The AI cloud-native inference suite (AI-ServingKit) is built based on the large-scale AI practices of ByteDance and the extensive business experience of ModelArk. It enables enterprises to run mainstream inference models efficiently, stably, and cost-effectively on large-scale GPU clusters through container orchestration, AI gateways, end-to-end inference observability, and distributed caching.

Leveraging features such as non-intrusive SDKs, automatic framework instrumentation, automatically generated TraceIDs, and preset dashboards and alert rules that fully cover the end-to-end path from inference engines to GPUs, APMPlus provides end-to-end observability in the AI-ServingKit solution. It collects metrics for APIG instances, inference engines, GPUs, and other resources to detect SLO metric fluctuations in a timely manner. Compared with traditional observability, end-to-end observability reduces integration time by 80 percent and reduces troubleshooting time by 70 percent. This achieves transparent end-to-end instrumentation and minute-level inference issue localization.

LLM end-to-end monitoring

APMPlus features transparent integration without instrumentation, compatibility with mainstream open-source frameworks, and end-to-end monitoring for self-managed inference scenarios. It provides a low-cost observability solution for LLMs and supports end-to-end monitoring for LLMs across various deployment modes.

Inference fault detection

By promptly tracking the performance of agents and AI inference scenarios through metrics, APMPlus can identify issues within one minute and locate them within five minutes.

Model metric monitoring

APMPlus fully collects statistics on relevant metrics in AI scenarios by model.

Agent topology map

APMPlus displays the upstream and downstream call relationships of the current AI business in real time and analyzes hotspots and errors.

Intelligent root cause localization

APMPlus supports multiple intelligent algorithms for quick metric diagnosis and identification.

Application evaluation system

APMPlus builds an objective scoring system for application running status based on backend system metrics. This helps you monitor the running status of each application system in real time.

Issue localization

APMPlus provides end-to-end issue detection and localization for applications in various forms.

Real-time metric tracking

APMPlus tracks metrics in real time before and after service releases. Comprehensive and real-time metrics help enterprises analyze data differences between versions.

Last updated: 2026.09.03 11:22:42