Mastering Online Business: How to Map Architecture, Build Multi‑Dimensional Monitoring, and Visualize Data

In operations work, understanding the full business architecture is essential for pinpointing single points of failure and hidden risks. The presenter, a senior operations manager at Meitu, outlines three key steps: mapping the architecture, establishing multi‑dimensional monitoring coverage, and visualizing collected data.

Mapping Business Architecture

From a user’s perspective, services appear as a seamless cloud, but operators must know how users reach the service and how the backend processes requests. Tools such as Firebug, Chrome DevTools, HttpWatch, and Charles help capture entry points, API calls, loaded resources, IP addresses, and response times. Commands like nslookup, dig, and ping verify domain IPs and detect hijacking.

On the server side, high‑availability and load‑balancing devices (e.g., F5, LVS) distribute traffic across multiple nodes. Operators can use LVS utilities to identify real servers handling specific client requests.

Understanding backend operations involves inspecting Nginx or HAProxy configurations, checking listening ports, running processes, and using strace to trace execution paths and locate configuration files. The /proc/<pid> filesystem provides detailed process information, while Nginx upstream definitions reveal service deployment details.

Different services may run on Tomcat, PHP, or other platforms. Operators should examine TCP/UDP connections, use ss to view inter‑service dependencies, and map master‑slave relationships for both application clusters and databases. This comprehensive view enables rapid fault isolation.

Multi‑Dimensional Monitoring Coverage

Effective monitoring starts with a clear understanding of business logic and module boundaries. Two approaches help create an up‑to‑date architecture diagram: manually drawing with Visio‑like tools (higher maintenance cost) or using scriptable graph languages such as DOT to generate diagrams automatically.

With an accurate diagram, operators can systematically verify monitoring coverage, ensuring each server and service node is observed. Monitoring should span three dimensions:

Basic monitoring at network and OS levels (e.g., Zabbix, Falcon).

Log monitoring using ELK stack to collect, filter, and report error logs.

Semantic monitoring that actively probes services (e.g., curl requests) via shell scripts or code in Python, PHP, etc.

Data Visualization

Monitoring generates abundant metrics: CPU, memory, I/O, API response times, request volumes, Nginx logs, and backend service logs. Consolidating these metrics onto a unified dashboard provides a holistic view of service health and traffic patterns. Custom dashboards can illustrate user distribution, regional performance, and carrier‑specific issues, enabling data‑driven decision making.

Conclusion

When taking over a new service, the first step is to understand its architecture, then implement comprehensive, multi‑dimensional monitoring, and finally present the data in clear visualizations. This systematic approach improves service reliability and operational efficiency, delivering better outcomes with less effort.