From Telecom to Startup: A Veteran Ops Engineer Shares Career Lessons

Guest Introduction

Wang Jinyin He previously worked at Tencent, YY, UC and runs the WeChat subscription account "Internet Operations Talk", which receives tens of thousands of views per article. In late 2015 he founded Youwei Technology, quickly attracting industry attention.

Opening Remarks

He originally wanted to talk about name services to match the tone of the KVM group, but was asked to share his operations experience instead.

He thanks the host and notes he will not discuss his neighbor's work.

Overview of Wang's Experience

He outlines his career in several stages, each with its own responsibilities and lessons.

1. Early Stage (Guangdong Puxin Technology)

After graduate school he joined a telecom service provider to develop a next‑generation BOSS system, participating in a telecom resource management system (simpler than CMDB) and eventually leading a team. He also took part in several system cut‑over projects.

Standardization awareness – telecom development follows strict processes from requirements to database design, using naming conventions like tb for tables and tf for fields.

Methodology cultivation – influenced by eTOM, NGOSS, domain‑driven design, etc.

2. Accumulation Stage (Tencent)

At Tencent he started with release automation and monitoring, then moved to on‑call duty, requirements, and later led the storage operations team.

Key takeaways:

Understanding the whole operations lifecycle from ITIL to tool‑based automation, covering basic, business, and storage operations.

Grasping the full internet technology stack by working in both front‑end and storage groups.

Collaborating with a high‑performing five‑person web‑ops team that handed over nearly 7,000 servers.

Recognizing that hardship drives growth; painful incidents become valuable lessons.

3. Release Stage (YY and UC)

After leaving Tencent he joined YY and later UC, applying Tencent’s ops practices to new environments.

Highlights:

Do not over‑rely on ops‑R&D many needs can be met by the ops team itself, unlocking hidden potential.

Ops‑R&D capability is a core competitive advantage.

Integrating development, testing, and operations (DevOps) yields better results than isolated tooling.

Implementing a name‑service center across companies.

He also contributed articles on Zookeeper‑based name service and gave talks on “Design and Implementation of a Full‑Stack DevOps Platform”.

4. Entrepreneurship Stage (Youwei)

He founded Youwei to replace traditional ITIL with an internet‑centric ops solution, focusing on automation (ITOA) and analytics (ITOA), with a shared metadata layer (CMDB).

Product highlights include end‑to‑end monitoring via name service, service tagging, and automated testing.

He invites engineers to join the Youwei R&D centers in Guangzhou and Shenzhen.

Good News! Resources

Presentation slides: Operations Era – Design and Implementation of a Full‑Stack DevOps Platform

http://pan.baidu.com/s/1bouculx

Q&A

Q1: How do you learn the vast amount of ops technology?

Answer: Treat failures as teachers, dig deep into root causes; explore knowledge trees (e.g., CPU metrics); exchange with experts; avoid positioning yourself as a tech guru in business ops.

Q2: Advice on building an ops team?

Answer: Align team structure with business scale and infrastructure; adopt quality/cost/efficiency/security metrics; consider diverse team composition.

Q3: How to start standardizing ops processes?

Answer: Identify the hierarchy of IT objects (hardware, OS, components, applications, scheduling) and define standards at each level.

Q4: How to handle pain points uncovered by incidents?

Answer: Turn problems into architectural optimization opportunities; use them to address technical debt.

Q5: Top‑down or bottom‑up when designing ops architecture?

Answer: Start with top‑down principles, then implement bottom‑up modules.

Q6: Should an existing ops system be rebuilt or iteratively improved?

Answer: Continuous iteration; the architecture must adapt to changing business scale and complexity.

Q7: How to standardize workflows like continuous deployment?

Answer: Model each object (e.g., DNS) with CRUD operations and map change scenarios (version upgrade, rollback) to automated processes.

GOPS 2016 Global Ops Conference – Shenzhen

Join the event on March 25‑26; details and registration are available via the linked image.

http://gops2016-shenzhen.eventdove.com/

Wang will present “Lean Ops System for High‑Performance IT”.