Region build as a service: Oracle’s unique approach to cloud expansion

By Jae Sook Evans

To support our global customers, we have been actively launching highly available cloud data centres globally that prioritise security, availability, performance, and price. These data centres are grouped into independent regions and each region offers hundreds of services (such as virtual machines and storage).

Towards the end of the last calendar year, we launched a total of eight regions – seven commercial regions (Singapore, Marseille, Israel, Abu Dhabi, Stockholm, Milan including our latest launch in South Africa), along with a second dedicated region for Nomura Research Institute in Japan – all within a 11-week period. That’s an average of one region every one-and-a-half week!

Most impressively, we’ve gained velocity at scale – not only have we reduced the time needed for each region build, but we’ve also more than tripled the number of services offered by Oracle Cloud Infrastructure (OCI).

Why are we building so many regions?

To meet our continued phenomenal growth, we’re building cloud regions as fast as customers around the world demand. Our customers have never been more dependent on the cloud than they are today. 

They need a cloud that’s architected to run their critical workloads and that supports their business needs with superior price performance, doesn’t overcharge for data egress, and is available in regions that meet their data locality requirements and internal governance needs. This also includes dedicated regions cloud@customer where customers get all the benefits of OCI cloud services maintained by Oracle, but inside their own data centres.  

What we changed

To meet this demand, we are treating our region builds “as-a-service,” and we’re turning this rapid and reliable capability into a core competency -- and a key differentiator. It is only by fully automating our builds and launches that we can reliably and rapidly launch all services in all regions, in any of the configurations we offer (commercial, government, dedicated region, etc.).

We invested in an end-to-end automation program that organises work and reports progress. Most importantly, we aligned on a set of measurable goals and outcomes as our success criteria. 

The new deployment technology in which we invested allowed us to dramatically reduce the time and effort it takes to build regions, as well as to create a uniform deployment pipeline across all region types. We used existing OCI engineering and architecture review processes to help connect teams with experienced architects and obtain feedback on the design and architectural changes that were necessary to reach region build automation. 

• Automating the build process end-to-end allows us to provide consistent services in a repeatable manner across all our regions, including commercial, government, and dedicated regions: Everything Everywhere.
• It also makes us much more efficient, as fewer engineering hours are spent doing manual tasks.

The devil and the details

Before our automation campaign, building out the OCI cloud consisted of carrying out thousands of infrastructure and software deployments, executed and coordinated by hundreds of engineers across many cross-functional teams. Without automation, this requires extensive manual coordination across many services and their dependencies in proper order and tedious manual data transfer processes.

Foundationally, we built our own automated deployment pipeline software and capabilities to:

• help identify service dependencies;
• build our cloud in the correct order computed dynamically as dependencies become available;
• reduce errors caused by change with configuration validation and pre-flight-checks;
• achieve full integration with our change management system; and
• configure and manage the state of our global fleets

From there, we extended the automation capabilities across all software components and the integrations between services that make up building out a region.

Lessons Learned

Automation can help businesses overhaul processes and achieve big savings – whether that be time, labour, or money – but implementing it requires iteration and experimentation. Our automation work taught us to:

• Have an automation test lab that runs continuously to mirror the operating requirements of a production region. This will ensure that building regions is fast, repeatable, and error-free.
• Manage your region build-as-a-service programmatically, with clearly defined goals and outcomes communicated to all levels of the organisation.
• Ensure all OCI services use the same platform and language to define their infrastructure and software configurations.

Automation is helping us build regions more quickly while slashing error rates, and as our teams achieve complete touchless automation, our engineers spend less time on tedious configurations and more time figuring out how to build an even better, faster cloud.

(The writer is the Chief Information Officer at Oracle)