How ‘泡姆泡姆’ Leverages Cloud‑Native Architecture for Global Low‑Latency Gaming

Game Overview

“泡姆泡姆” is a multiplayer cooperative party‑adventure game by Eagle Corner Network that blends colorful shooting, match‑3, physics interaction and puzzle solving, with additional arcade mini‑games. It supports keyboard/mouse or controller, offers local two‑player co‑op and online party modes for 3‑4 players.

Cloud‑Native Architecture

The game runs on a cloud‑native stack built on Alibaba Cloud Container Service (ACK) with OpenKruiseGame (OKG) for fine‑grained workload governance. The architecture is distributed, highly available, horizontally scalable and fully observable, enabling global deployment across 4+7 regional data centers.

Elastic Room Management

Using Keda and custom OKG triggers, the system automatically scales room servers based on player load, maintaining a minimum pool and provisioning additional instances within seconds during peaks. Custom service‑quality features allow intelligent room lifecycle management and resource reclamation during low‑traffic periods.

Zero‑Downtime Versioning

RoomManager acts as a version‑control hub, routing players to appropriate server versions. Multi‑version routing and gradual rollout enable seamless updates without server downtime, with old instances retiring after ongoing matches.

Observability Stack

A three‑layer observability platform combines SLS log collection, CloudMonitor metrics, and ARMS/OpenTelemetry tracing. Logs from 11 regions are stored locally and aggregated for global queries; metrics cover CPU, memory, network, I/O, and business KPIs; tracing provides full‑link diagnostics, P99 latency analysis and service topology.

Monitoring and Real‑Time Insights

CloudMonitor, integrated with Grafana, offers unified dashboards for global resource health, showing container scheduling pressure, database slow‑query trends, and network latency. The system correlates business metrics (online users, match time) with infrastructure metrics to provide proactive alerts and capacity planning.

Recent Upgrades

Tracing has been migrated to OpenTelemetry, reducing storage and operational costs by 90% and eliminating the self‑built Jaeger cluster. ARMS now delivers service topology, latency analysis and root‑cause diagnostics across regions.

Future Directions

Plans include AI‑driven anomaly prediction, automatic root‑cause recommendation and operation intelligence, aiming to further reduce downtime and enhance player experience.