Key results:

• $25K Monthly Savings: Switching from N2 to N2D for spot GKE node pools.
• $13K Monthly Savings: Cloud Storage cost reduction through deep object-level access analysis.
• Increased Engineering Productivity: Equivalent to saving the work of a full-time engineer.

About Ninja Van Group

Ninja Van Group is a tech-enabled logistics organisation, backed by marquee investors including GeoPost, Alibaba Group, and B Capital Group. A decade in operations strengthened Ninja Van Group’s e-commerce express network, enabling diversification across e-commerce and express logistics. Ninja Van Group leverages innovative technology to provide seamless delivery solutions and optimise efficiency for businesses of all sizes. Headquartered in Singapore, Ninja Van Group operates in six countries: Singapore, Malaysia, the Philippines, Indonesia, Thailand, and Vietnam.

According to the GCP case study, Ninja Van Group’s DevOps team relies on Google Kubernetes Engine (GKE) to autoscale node pools dynamically instead of using vertical scaling, which could lead to downtime. By setting minimum and maximum deployment limits based on CPU utilization, the team can continuously monitor the Kubernetes cluster, ensuring that more pods are assigned when necessary.

“Thanks to Horizontal Pod Autoscaler, we can dynamically adjust our workloads and scale down to reduce costs when resources are no longer needed,” says Shaun. “During peak demand, Ninja Van Group’s website and mobile applications can handle 10 times the normal traffic without breaking a sweat.”

Since delivery service demand varies by country, Ninja Van Group leverages historical data stored in Cloud Storage for capacity planning. For example, analyzing past delivery orders and driver efficiency enables the company to hire additional drivers during holiday seasons to ensure timely deliveries.

“We collect hundreds of gigabytes of data daily, from driver location data captured through the Ninja Driver app to click tracking on our website,” says Shaun. “We use Spark jobs to analyze this data in Cloud Storage, helping us forecast manpower demand.”

GKE Optimization

Following the integration with Rabbit, multiple recommendations were made for optimizing GKE node pools. One key improvement was switching machine types from N2 (Intel CPUs) to N2D (AMD CPUs), which are 13% cheaper and offer 20% better performance. Since Ninja Van Group had a large spot node pool without any commitment, this adjustment resulted in monthly savings of $17K when we compared one-by-one. Additionally, improved performance allowed the company to operate with fewer nodes, increasing total savings to $25K per month.

Another $20K monthly saving was achieved by switching N2 to N2D. However, Rabbit carefully highlighted that Ninja Van Group had a three-year active resource-based commitment, allowing the team to plan further optimization strategies around the expiration of this commitment.

At the workload level, Rabbit provided recommendations for optimizing CPU and memory allocation. Ninja Van Group set target utilization settings to fine-tune how recommendations were calculated. For example:

• Development Environment: 20% buffer above P99 utilization.
• Production Environment: 40% buffer above P99 utilization.

These adjustments led to an additional monthly savings. However, implementing them manually was impractical, as over 1,800 workloads required adjustments. To overcome this, Ninja Van Group utilized Rabbit’s Automation feature, deploying it first in development and, after rigorous testing, in production. This significantly reduced engineering time and accelerated cost-saving implementation.

Cloud Storage Optimization

With petabytes of stored data and millions of objects, managing Cloud Storage efficiently is a challenge. Native GCP tools offer limited functionality for large-scale bucket management. Rabbit processed terabytes of object access logs to identify and isolate unused objects, significantly accelerating the cleanup process. As a result, Ninja Van Group achieved a $13K monthly saving on storage costs.

Boosting Engineering Efficiency

By automating GKE optimization and providing deep insights into GKE and GCS cost data, Rabbit effectively saved the equivalent of a full-time engineering team’s workload. Additionally, Rabbit’s cost anomaly alerts, weekly and monthly reports, and notifications for new optimization opportunities ensured the team remained in control of their expenses while maximizing efficiency.

Conclusion

With Rabbit’s automated optimization strategies, Ninja Van Group successfully reduced costs by a monthly $83K, enhanced system efficiency, and improved engineering productivity. By leveraging data-driven insights and automation, the company continues to scale effectively while maintaining cost efficiency across its cloud infrastructure.