Cloud Computing in Asia: A 2015 Prediction Revisited
Looking back at an early view of Alibaba Cloud, regional infrastructure and why cloud architecture is never geography-free.
Archive note: Originally published on Mullener in 2015. This edition removes some period-specific market speculation and concentrates on the architectural point that remains relevant.
In 2015 I was working across public cloud and VMware-based environments. One challenge was providing compute and network connectivity between China, the wider Asia-Pacific region and the UK.
It was easy to describe cloud as compute capacity available anywhere. The actual system was shaped by geography: transit providers, latency, local regulation, security controls, service availability and the way each carrier operated.
The emerging regional cloud
AWS was already mature, Google was expanding its compute platform, and Alibaba was becoming a serious cloud provider in China. My view at the time was that a strong regional provider could solve problems that were awkward for Western platforms to solve from outside the region.
That was not simply a question of server location. A provider with local infrastructure, commercial relationships and operational knowledge could offer a different path through:
- domestic and international network transit;
- regional performance and latency;
- local regulatory requirements;
- technical support in the same market;
- partnerships with existing virtualisation and infrastructure companies.
The architectural lesson
Cloud resources may be ordered through a global console, but their behaviour is never geography-free. A design that works well between London and Dublin cannot be assumed to behave the same way between London and mainland China.
Before choosing a provider or region, the questions should include:
- Where are the users and data located?
- Which networks carry the traffic in each direction?
- What are the latency, loss and bandwidth characteristics?
- Which services are actually available in-region?
- What regulatory or data-residency constraints apply?
- How will the system be operated when a cross-border link is impaired?
Looking back
The names, products and market positions have evolved since 2015. The core prediction was sound: Asia-Pacific cloud infrastructure would not merely be a copy of the US market, and regional providers would matter.
The more useful conclusion today is architectural rather than competitive. “The cloud” is a collection of physical regions, networks and legal jurisdictions. Good platform engineering begins by making those boundaries visible.