How Gmarket replaced Control Center and cut Kafka costs with Kpow
Challenge
After adopting the licensed Confluent Kafka distribution, Gmarket re-evaluated the cost of that commercial license and decided to migrate to community Kafka. That move meant losing Control Center, and the team needed a replacement that covered its most critical monitoring needs: consumer lag and topic offset visibility.
Solution
Kpow gave Gmarket:
- Consumer lag and topic offset visibility to replace Control Center
- Data Inspect, letting teams browse and filter partitions without a consumer
- Self-service reassignment, offset resets, and topic operations for non-Kafka engineers
Result
- A more cost-effective alternative to Confluent's licensed Control Center
- Roughly 150 users across production and development, spanning 10 Kafka clusters
- Reduced operational costs, with non-Kafka engineers now handling tasks that once needed specialist support
“Features such as reassignment, offset resets, topic operations, and consumer state monitoring are all available through the UI, which enables even non-Kafka engineers to perform operational tasks easily and efficiently.”
Yubin Kim, Engineer, Gmarket
Moving off a licensed Kafka distribution
Gmarket is one of South Korea’s leading e-commerce marketplaces, founded in 2000 and now part of Shinsegae Group. As of October 2025, the platform ranks fifth in the country by monthly active users, and distributes products from 20 million Korean sellers to a further 160 million customers across Southeast Asia through its Lazada partnership. When the company first adopted Kafka, it ran the licensed Confluent distribution, which bundled in Control Center for cluster monitoring. Over time, Gmarket re-evaluated whether it needed the commercial license at all, and decided to migrate to community Kafka. That decision meant giving up Control Center, and the team needed something to take its place.
For Gmarket, the priority was narrow and clear: consumer lag and topic offset visibility, the metrics its teams relied on most to know their Kafka estate was healthy. The team compared several open-source options, including Kafka UI and CMAK, before settling on Factor House. As Yubin Kim, an engineer at Gmarket, put it, Kpow was “a more reasonable and cost-effective choice for our needs.”
Inspecting partitions at scale, without writing a consumer
Kpow is now in use across roughly 150 users spanning Gmarket’s production and development environments, running across 10 Kafka clusters, licensed on a per-cluster basis. Of everything Kpow offers, Yubin singled out Data Inspect as the feature the team would miss most. It lets Gmarket’s clients browse partitions and apply filters directly, without needing to write and run a consumer of their own. “This is the most frequently used feature across our teams,” Yubin said.
Lower costs and broader access to Kafka operations
The biggest impact for Gmarket has been reduced operational costs. Tasks like reassignment, offset resets, topic operations, and consumer state monitoring are all available directly through Kpow’s UI, which means they’re no longer bottlenecked on specialist Kafka engineers. As Yubin put it: “Features such as reassignment, offset resets, topic operations, and consumer state monitoring are all available through the UI, which enables even non-Kafka engineers to perform operational tasks easily and efficiently.”
Together, moving off a licensed Kafka distribution and onto Kpow has let Gmarket keep the monitoring visibility it depended on, at a more reasonable cost, while opening up day-to-day Kafka operations to a much wider group of engineers than before.
