How Six Pillars Transform Data Center Operations into Full Automation

Editor

Old Bian (content collection & article compilation)

Guest Introduction

Zhi Jin – real full name (note: the surname Zhi is rare).

2006‑2011: Responsible for system operations at Alipay, one of the first teams in Alibaba Group to build an operations‑automation system.

2011‑2014: Managed operations tools and private‑cloud construction at the headquarters of China Construction Bank, applying Internet experience and open‑source software for secondary development.

Currently founder of Hangzhou Yunji Technology, dedicated to turning operations‑automation experience and DevOps ideas into products.

Topic Overview

Although terms like “operations automation”, “cloud computing” and “DevOps” are popular, many people’s understanding is narrow or even incorrect. In this session, guest Zhi Jin shares his experiences from large‑scale Internet companies and traditional enterprises, and, starting from the overall data‑center operations‑management system, presents viewpoints he has distilled from years of practice.

Share Transcript

During his time at Alipay, the business grew fourfold each year, expanding from 100 machines to tens of thousands in three years. Early automation was forced by rapid growth, a reactive process of discovering and fixing problems—an experience common to fast‑growing Internet firms.

Internet‑level operations automation can be described as “bottom‑up, wild growth”: quick to show results but hard to pause for reflection, akin to the Sword Sect of Mount Hua. After moving to the Construction Bank, I studied foreign best practices such as ITIL and COBIT, and evaluated products from IBM, HP, BMC. While the methodologies are powerful, their complexity makes implementation difficult, comparable to the Qi Sect of Mount Hua. Combining Internet practice with these insights, I eventually achieved a “sword‑and‑qi” dual cultivation, dramatically advancing my automation capabilities.

The focus now is on building a comprehensive data‑center operations‑management system. The following points are highlighted:

Many people mistakenly equate “operations automation” with simple “task automation”. Tools like Puppet are often seen merely as automation tools, whereas their core value lies in configuration management.

Cloud computing introduces low‑cost, standardized hardware and open‑source software, enabling self‑service and automated resource delivery. However, after implementation, infrastructure scale expands rapidly, increasing operational complexity.

Definition of Data‑Center Operations Management

Previously I introduced the concepts of “black‑box” and “white‑box” operations. To clarify, the International analysis firm Gartner summarizes data‑center operations as “I&O (Infrastructure & Operation)”, covering both infrastructure construction (pre‑deployment) and operational management (post‑deployment).

Automation is one aspect of operations. Broadly, automation means establishing closed‑loop systems. At the infrastructure layer, a small loop links operations and operations; at the operational‑management layer, a larger loop connects operations and development—essentially DevOps. The ultimate goal is full service‑orientation and full automation.

IT operations vs. IT service management differ in that the former is reactive, the latter proactive.

Note: The chart comparing private‑cloud O and I costs is the only image sourced from the web; it aligns with my understanding and was used without further verification.

Future Data‑Center Capabilities

The “six‑fold” summary captures the future capabilities of data centers.

Virtualization

Automation

Integration

Standardization

Visualization

Intelligence

Combining these six aspects constitutes comprehensive automation.

Three Core Elements of Operations

The essential elements are not servers, storage, or network, but data, processes, and actions. Orchestrating data and processes forms the core; actions are merely the implementation means.

Three Stages of Operations Automation and Closed‑Loop Process

The lower‑right quadrant shows the initial stage, where tools, processes, and data are tightly coupled, manageable only by small teams.

The lower‑left quadrant represents the intermediate stage: processes drive data, which in turn drives actions, forming a small closed loop.

The top‑most stage is the ideal architecture: a CMDB acts as the central control node, using data to govern all processes and tools.

This diagram illustrates the final closed‑loop process we aim to achieve.