Using KSQL to Reduce False Positives and Improve Observability At Scale

The real-time nature of an e-commerce business poses many challenges – namely, the ability to react quickly to resolve incidents affecting customer experience. To do that, you need to capture logs from various applications into a central system, then analyze them to improve observability and telemetry and reduce the number of false-positives for a substantially-improved NPS score. Using Apache Kafka and Confluent allows us to innovate with KSQL features like sliding windows and snapshots of time series data. In the long term, this will result in a trustworthy logging & observability system. In this talk we will share: ⁃ Our learnings leveraging KSQL at scale to perform real-time stream processing ⁃ Our journey to improve the overall skill level in our teams ⁃ The pros and cons of using KSQL at scale in telemetry and observability type use cases.

Hong Woo Lee
Senior Architect, ebayKorea