The lure of Google Cloud for Singapore’s Ninja Van

When regional logistics startup Ninja Van switched to Google Cloud from Amazon Web Services (AWS), it found that its applications were running faster than before.

“We were looking at our telemetry data for the same machines with the same amount of memory and CPUs, and realised our services were performing much better than when we were at AWS,” said Shaun Chong, cofounder and chief technology officer (CTO) at Ninja Van.

Chong and his team wondered if Google Cloud was using better processors that led to the performance improvements, but later concluded that it was Google’s network infrastructure that made the difference.

“Everything is all software defined in the datacentre,” Chong said. “Google can guarantee rack-to-rack latencies and capacities inside the datacentre – even between zones where we saw five to 10 times lower latencies. We basically got an almost free upgrade in performance.”

For a while now, Google Cloud has been touting its network performance as one of its key differentiators over rivals in a competitive public cloud services market and has been ramping up infrastructure investments across the Asia-Pacific region.

The cloud supplier is also part of a consortium that built the Indigo subsea cable that connects Australia and Southeast Asia in an effort to mitigate disruptions caused by fibre cuts, and to meet the growing demand for internet services in the region.

The fact that Google Cloud is part of the Indigo consortium was indicative of the cloud supplier’s efforts to deliver higher service levels. Rather than rely on third-party telcos to provide the links between its datacentres, Google Cloud would be able to do so on its own and get guaranteed bandwidth at a more predictable cost.

Having a more predictable operating expenditure was also one of the reasons behind Ninja Van’s decision to switch to Google Cloud. With AWS, Chong’s team was using reserved instances to save money, but had to decide how much resources they needed to buy in advance.

“The concept of reserved instances made it very difficult because our traffic today is very unpredictable,” Chong said. “We need to have the ability to scale up and down any time and while other cloud providers allowed us to do so, it’s not great in terms of cost optimisation.”

Chong said with Google Cloud, Ninja Van saw an immediate reduction in costs because of automatic sustained usage discounts, adding that costs can be brought down further with committed usage.

The move to Google Cloud took just a day, according to Chong. At the time, his team was using self-hosted Kubernetes, the popular open-source container orchestration platform, but decided to switch to Google Kubernetes Engine, a managed environment for deploying containerised applications easily, to alleviate the need to manage and test upgrades to the Kubernetes platform.

“We used to have three people managing self-hosted Kubernetes, and we just need one person today,” Chong said. “In fact, we can even get someone on board who doesn’t know Kubernetes to manage and create clusters very easily.”

“With just one command in the CLI [command line interface], we can bootstrap clusters literally within minutes. We used to write lots of scripts and configurations and had to figure out what virtual IPs we needed in the service and cloud layers. But today, everything is just a few clicks away.”

As a nimble startup, Ninja Van releases new features often, sometimes hundreds of times a day. To test new code on the fly, it started a new initiative called Project Gaia, named after the Greek goddess and creator of life, to create new environments for quality assurance (QA) testing. 

“We're spinning up full clusters in separate environments to allow our QA teams to test product features independently, and GKE is helping us to do that very efficiently,” Chong said.

Although Ninja Van is looking to consolidate all its compute resources on Google Cloud, it still uses AWS for some storage resources for now, as well as the Amazon Route 53 service to route traffic to China. “We also use Alibaba Cloud and Tencent Cloud but that’s just for our China business,” Chong said.

According to a Boston Consulting Group study that covered six countries in Asia-Pacific, public cloud adoption in the region is expected to contribute $450bn from 2019 to 2023 and create 425,000 direct jobs and another 1.2 million jobs indirectly.