AntMonitor: Evolution, Features, and Core Technologies of Ant Group’s Observability Platform

Introduction : At the inaugural "Stability Assurance Plan" cloud system stability conference hosted by the China Academy of Information and Communications Technology, Ant Group’s AntMonitor platform received the highest "Advanced" certification for observability capabilities.

1. Development History : Starting with an early monitoring platform before 2011, Ant Group built a full‑stack observability system through successive phases—initial monitoring, business‑centric monitoring (2012‑2017), and post‑2017 holographic, data‑driven, and AI‑enabled capabilities, achieving one‑stop monitoring across client, server, business, and infrastructure layers.

2. Featured Product Capabilities :

Holographic Observability: Unified collection of metrics, traces, logs, and performance analysis, breaking data silos and enabling end‑to‑end visibility.

Integrated Performance Analysis: Fine‑grained CPU flame‑graph analysis from macro metrics down to specific code lines.

Efficient Integration Model: Standardized, entity‑based, and topology‑aware modeling to simplify onboarding of heterogeneous monitoring entities.

Built‑in Data Intelligence: Real‑time data feeds power AIOps, supporting PromQL and SQL queries over both time‑series and dimensional tables.

Algorithm Engineering Platform: End‑to‑end pipeline for model deployment, training, regression, and data labeling, enabling intelligent risk detection.

Monitoring‑as‑a‑Service (MaaS): Exposes monitoring compute, storage, algorithms, and visualizations as services for SRE teams, promoting reusable analysis capabilities.

3. Core Platform Technologies :

Fusion Time‑Series Data Platform (Pontus): A unified CMDB‑plus‑time‑series solution handling millions of tables and billions of data points.

Data Management: Comprehensive lifecycle handling—collection, computation, storage, and consumption—comparable to AWS Timestream or Azure Time Series Insights.

Multi‑Dimensional Time‑Series Model: Snowflake‑style schema separating dimensional (metadata) and time‑series tables for flexible querying.

Massive Real‑Time Processing Architecture: Regional multi‑active design processing ~40 TB/min and 200 billion points per minute, with operator push‑down and near‑edge computation.

High‑Performance Time‑Series Database (CeresDB): Designed for ultra‑high write/read throughput, high availability, multi‑tenant control, and seamless integration of time‑series and analytical workloads.

New Hardware Exploration (AEP): Leveraging App‑Direct persistent memory to bridge the performance gap between DRAM and SSD, reducing query latency for hot data.

Conclusion and Outlook : AntMonitor’s continuous evolution, open‑source ambitions, and commercialization of observability components aim to provide a stable, scalable foundation for digital transformation across industries, with plans to open‑source CeresDB and extend AI‑driven monitoring services.