Designing the Blue Whale Ops Platform: Architecture, PaaS, and Automation Insights

Speaker Introduction

Dang Shouhui Since 2012 he has been responsible for designing, building, and operating Tencent Game's operations support system (Blue Whale), integrating SOA, cloud, and big‑data concepts to create an independent ops infrastructure and promote DevOps within the industry.

Talk Overview

Blue Whale System Design Thoughts and Architecture Analysis

1. Positioning

Although initially marketed as an ops solution, Blue Whale was never limited to a single point; it also serves business operations and team management.

An enterprise must focus on five aspects: product design, product development, market channels, business operations, and team management. Blue Whale targets the latter two.

These are realized through various operational systems rather than abstract processes.

Key Takeaway

If viewed from the perspective of building enterprise operational systems, Blue Whale is a vertical PaaS; from a productization view, it is an enterprise operating system.

2. Design Philosophy

Unlike most companies that follow ITIL‑style incremental productization, Blue Whale adopts a distinct approach tailored to its unique business needs.

2.1 Enterprise Business Architecture

Enterprises may have multiple businesses; development teams often standardize frameworks or rely on public PaaS for fully managed services. Private clouds coexist with public clouds, making cross‑cloud management essential.

Because many games are developed externally, Blue Whale cannot embed itself into the codebase; instead, it aims for unattended operations without altering business architectures.

2.2 Operational Process Overview

Key processes such as release, fault handling, and configuration refresh are represented as point‑and‑line diagrams, emphasizing that any operation composed of connected steps can be adapted by Blue Whale.

2.3 Blue Whale Job Platform

The platform registers scripts, distributes execution across thousands of servers, supports file transfer, and standardizes script management, turning ad‑hoc root commands into reusable atomic tasks.

2.4 Script Modularity and Reuse

By backing up scripts, tracking execution history, and exposing them via APIs, the platform enables high‑concurrency, high‑availability execution without requiring developers to handle low‑level details.

2.5 Blue Whale Configuration Platform

It stores foundational data and business topology, integrates with CMDB, and allows custom attributes per business, facilitating cross‑cloud IP conflict resolution and automatic environment synchronization.

2.6 Blue Whale Control Platform

Provides unified agent‑based control over IaaS resources, supporting massive global and cross‑cloud management, as well as large‑scale data collection.

Automation Stages

Automation evolves through four stages:

Script Automation : manual scripts executed by operators.

System Automation : web‑based tools expose functionality via UI.

Scheduling Automation : APIs enable cross‑system orchestration with a central scheduler.

Intelligence : fully unattended workflows with dozens of linked nodes.

3. PaaS Perspective

PaaS requires full code hosting without ops overhead and provision of cloud APIs; its main advantage is low cost.

3.1 Building an Operational System

Traditional development involves environment setup, code deployment, monitoring, and logging. With PaaS, developers focus solely on application logic, leaving deployment and scaling to the platform.

3.2 Front‑end and Back‑end Enablement

Pre‑built UI components and backend services (authentication, logging, cloud APIs) allow operators to create functional tools with minimal coding.

3.3 Operations Data Platform

The data platform, powered by a single Blue Whale agent, collects, aggregates, and processes large‑scale operational data. It supports YAML‑defined parsing, SQL engines, and visual drag‑and‑drop dashboards, reducing reliance on custom big‑data pipelines.

Future Topics

Further sessions will cover advanced SaaS implementations and self‑healing fault mechanisms.