Managing large volumes of log data, minimising duplicate information and reducing latency were key challenges faced by Nexon. As one of the world's leading game developers and publishers, ensuring that data is moved quickly and efficiently is critical for ensuring the end user experience is optimised. The company boasts popular gaming titles including MapleStory, ArcRiders, and Dungeon & Fighter. But the company found that the solutions they had in place were not delivering the level of service they demanded.
That led the company to a significant shift from their previous platform. While they were using Apache Kafka through a cloud hyperscaler, they found performance was not meeting the company's needs.
Nexon's Minwoo Kim, Team Manager of the Promotion Team, explains.
"We decided to migrate to Confluent Cloud because we wanted a fully automated, SaaS-based managed solution. In terms of management we had confidence that Confluent, as the creator of Kafka, would deliver better than anyone else from both a technical and operational standpoint."
For Nexon, the most critical requirements for its data pipeline were reducing log transmission latency and ensuring exactly-once delivery semantics.
Kim says that after migrating to Confluent Cloud metrics such as network jitter became much more stable when collecting data across global regions. Before the migration to Confluent, Nexon used valuable resources to perform tuning tasks to keep service brokers running reliably. After the transition, it no longer needed to worry about that. Network stability drastically improved and duplicate records in logs have been significantly reduced.
The Confluent Cloud cluster architecture deployed by Nexon enables client access from both inside and outside the cluster. Internal components for log processing are directly connected within the cluster to reduce latency and traffic costs, while external components use fixed IP addresses through a Kafka proxy to ensure network security.
By configuring a single cluster to support both internal and external connectivity, Nexon was able to achieve signfiicant cost savings by eliminating redundant data replication.
One of the most critical peak periods for Nexon is the launch of a new game. Everything must work perfectly. Nexon can leverage the compression capabilities provided by the Kafka Producer. Logs are text-based with many redundant elements, so compression delivers a 70-80% reduction in traffic costs.
"When ingesting data into our data lake, we employ various buffering strategies to batch records together rather than processing them individually, which further reduces costs. Beyond this, we have plans to implement Apache Avro, and we anticipate additional cost savings from that initiative," says Kim.
Management of the technology has been optimised through the Confluent Cloud Console. The dashboard clearly displays essential metrics for cluster health enabling Nexon to quickly identify system anomalies and explicitly see items requiring action.
"The ability to view the status of each client by service account has also proven helpful," adds Kim.
The increased visibility not only helps Nexon to find faults faster and optimise performance. By eliminating data duplication and loss, it has built greater confidence in collected data which it has been able to leverage for marketing and operational initiatives.
The new platform enables Nexon to receive in-game logs from multiple games across global, multi-platform environments in real-time without data loss or duplication. Using proprietary monitoring tools, the company is deliver real-time analytics and in-game events.
Kim explains.
"Depending on the nature of the data, we employ a selective operational strategy. For data where speed is critical, we configure dedicated pipelines to reduce latency. And since we operate across multiple regions, we pay close attention to optimising data transfer speeds between regions and managing associated costs."
Nexon's migration to Confluent Cloud demonstrates how a purpose‑built, fully managed Kafka offering can turn a high‑volume, latency‑sensitive game‑log pipeline into a resilient, cost‑effective system. By consolidating internal and external traffic, eliminating duplicate records, and leveraging compression and batching, the company has optimised network performance, reduced network traffic costs and has real‑time analytics across dozens of global titles. This is through a single, fully monitored cluster that frees engineering teams to focus on game innovation rather than data‑pipeline management.