Inside the Singapore government’s cloud journey

Ensuring resiliency

Chan said having transparency would enable the government to ensure the resiliency of mission-critical systems across availability zones – as well as their survivability. “What happens if there’s a shutdown and the zones are disconnected from the internet? How long will they survive and what are the consequences?”

But such criteria for hosting mission-critical systems would drive up costs significantly and deny the government of the benefits of a rich ecosystem of third-party services available through the public cloud.

With a wish to have its cake and eat it, the government approached the major hyperscalers for a service that would provide it with the transparency and higher levels of security it needed to host mission-critical systems without compromising the benefits of cloud.

Chan said that while there were sovereign cloud offerings in the market, “everything would be locked up” so there was no offering that met the government’s requirements. “We did workshops and deep-dived into things like security caging and dedicated personnel with the hyperscalers – AWS was most ready for the job and that was how Dedicated Local Zones was conceived,” he said.

With Dedicated Local Zones, Chan said the government now offers public sector agencies three cloud deployment options – commercial cloud services for quick deployment of workloads; GCC for those concerned about localisation; and Dedicated Local Zones for highly sensitive and mission-critical workloads.

Besides AWS, the Singapore government also uses public cloud services from Microsoft Azure and Google Cloud. Chan said the government will evaluate other hyperscaler offerings similar to Dedicated Local Zones if they become more mature.

“From a safety and resiliency point of view, we don’t want to put all our eggs in one basket,” he said. “The design philosophy of each cloud provider can be quite different, and it’s not about whether one is better than the other. We’re working with all of them, and when a product comes along that makes sense, we’ll look at it.”

But running workloads on multiple public cloud providers can be daunting as it requires technology teams to be skilled in the inner workings of each cloud. While the government doesn’t dictate the choice of cloud provider for agencies, it has called for cloud providers to invest in training, said Chan.

AWS, for example, has trained over 200,000 individuals in Singapore on cloud skills since 2017 through various programmes such as Skill Builder, as well as re/Start, which prepares individuals for careers in the cloud. Earlier this year, it partnered with Ngee Ann Polytechnic to provide over 500 ICT students with access to its training and certification programmes over three years.

Cloud migration approaches

The Singapore government has taken pragmatic approaches to its cloud migration efforts. These include lifting and shifting workloads from on-premise environments; re-platforming, such as compiling code from legacy platforms like Solaris into Java; replacing legacy systems with software-as-a-service alternatives; rearchitecting into cloud-native applications; and eliminating workloads that are hardly used.

Chan said of all the approaches, the lift-and-shift approach was most criticised as it was perceived to be of little value. But he defended the approach, noting that lifting and shifting legacy workloads gives agencies and their suppliers time to learn how to work with cloud technologies.

“Also, you can’t rearchitect all the applications to the cloud because it’s a lot of work and you don’t have the time and money,” he added. “Rearchitecting can happen when an application reaches its end of life, and the faster you move the better because it allows you to rationalise your infrastructure.”

The archetype of each application is also a factor in determining which migration approach to take. Chan said customer-facing applications tend to be rearchitected because of the need to make them more scalable, elastic and customer-friendly using cloud-native services and application programming interfaces that provide access to third-party services such as payment systems.

Just as important is the need for applications to failover to a different availability zone during a datacentre outage. He said that while it’s much easier to do this with the cloud today with multiple availability zones, IT teams will need to know how to configure applications properly for the failover to occur.

“The technology might be there, but the way it is used is important,” said Chan. “That’s why training is so important, not just for students but also our vendors. If we don’t upgrade the industry fast enough, we will have problems.”