What Drives China Mobile’s Cloud Computing Strategy? Insights and Lessons

China Mobile Cloud Computing Vision

Since 2011 China Mobile has pursued a cloud strategy that separates private and public clouds to lower construction and operation costs, improve information processing, enable service innovation, and strengthen core competitiveness.

Private Cloud: Enterprise‑wide resource pools (supporting clouds and business clouds) that provide elastic, multi‑tenant resource allocation.

Public Cloud: Standardized resources for small‑to‑medium enterprises and individual users, similar to Amazon AWS, and customized platforms for large enterprise customers.

Core Cloud Computing Technologies

The most critical attributes of cloud computing include elastic scaling, distributed resource management, and lightweight virtualization. Achieving true “big cloud” requires:

Resource management via virtualization (e.g., LXC/Docker) and distributed scheduling.

Distributed coordination services for unified naming and stateless application deployment.

Key Technical Elements

Elastic scaling hinges on resource management and scheduling; lightweight virtualization (Docker/LXC) replaces heavyweight VMs; distributed coordination (e.g., ZooKeeper, Mesos, YARN) enables dynamic resource allocation; load balancing handles front‑end traffic.

Evolution of Cloud (The “Cloud Evolution Theory”)

True cloudification is not about shrinking clouds but expanding them—moving from “small cloud” (basic virtualization) to “big cloud” (data‑center‑level dynamic deployment). Many enterprises still confuse X86/VM deployment with genuine cloud transformation.

Reflections on Cloud Construction

Common pitfalls include siloed application‑centric cloud projects, low resource‑pool utilization, and fragmented management structures that defeat the economies of scale cloud promises. Standardized, platform‑centric approaches are needed instead of merely buying X86 servers and deploying VMware.

Zhejiang Mobile Cloud Development Thoughts

Four strategic directions are highlighted:

Achieve “big cloud” through breakthroughs in distributed scheduling and coordination.

Transition from IOE‑style architectures to open‑source solutions (e.g., PostgreSQL, MySQL) while managing technical debt.

Promote a layered service model: IaaS (small cloud), APaaS/TPaaS/DPaaS/BDPaaS (mid‑cloud), and EPaaS (enterprise‑level cloud operating system).

Standardize service catalogs to decouple technology implementation from application consumption.

Key Takeaways

X86 virtualization alone does not equal cloud; PaaS resource pooling and EPaaS components are essential for true “big cloud”.

Integration of big‑data platforms, lightweight containers, and orchestration tools (Hadoop, Docker, ZooKeeper, Mesos) will shape future cloud architectures.

Application‑centric “cloudification” is becoming obsolete; focus should shift to platform construction and migration.

Adopt a bottom‑up, platform‑centric view rather than top‑down, siloed implementations.

Resource pools must be unified across technology, management, and security domains, not fragmented by department.

Q&A Highlights

Q1: Why invest heavily in a “mid‑cloud”?

A: To address loss of core capabilities, siloed architectures, and low resource utilization, and to gradually move toward a true data‑center‑level cloud.

Q2: How does decoupling technology services from implementations work?

A: By exposing standardized interfaces (e.g., a cache service) so applications need not know whether the backend is Memcached, Redis, or another store.

Q3: How to handle capacity constraints before achieving a full “big cloud”?

A: Start with PaaS resource pools and platform standardization to improve utilization and agility, then incrementally evolve toward larger cloud capabilities.