“Our primary business challenge is maintaining seamless, high-quality gaming experiences across the globe,” Deng Ding, Director of Site Reliability Engineering, NetEase Games. “For real-time combat video games, the stakes are fundamentally different from casual games. Every frame matters. Every millisecond is a competitive advantage or disadvantage.”

Titles like FragPunk and Dunk City Dynasty, performance is inseparable from player experience. Matches depend on precise, real-time interactions between tens of thousands of players at once. No one likes lag, and even small delays can disrupt gameplay.

At the core of its architecture is Google Compute Engine, where FragPunk and Dunk City Dynasty runs performance-critical game servers on C4 virtual machines and central services on N4 virtual machines. These instances provide the compute power needed to support real-time combat scenarios, delivering consistent performance at 120+ FPS even under peak load.

To orchestrate these services at scale, the development team uses Google Kubernetes Engine (GKE). For example, during the launch of FragPunk, GKE enabled the team to automatically scale infrastructure in just five minutes—down from several hours—ensuring a seamless experience for more than 100,000 concurrent players.

Performance gains were immediate and measurable. By leveraging high-performance compute and Google’s global network, 2 titles reduced global response times by 67%, delivering the low-latency gameplay required for real-time competition. Players experienced smoother, more responsive matches across regions.

Improvement extended beyond performance. Resource utilization increased significantly, which enabled the project team to reduce server costs by 40% while maintaining higher levels of reliability.

Importantly, these performance and efficiency gains translated into measurable business impact. For Dunk City Dynasty, improved stability and responsiveness led to an 18% increase in Day-2 retention.