This article details ByteDance's end‑to‑end design and engineering of a low‑overhead, continuous profiling system that collects, processes, and analyzes resource usage across a million‑scale data center, enabling precise performance optimization and significant CPU cost savings.
This article details the end‑to‑end optimization of 58.com’s commercial traffic ranking system, covering data‑flow upgrades, advanced feature engineering, real‑time and multi‑task model improvements, and a multi‑factor ranking mechanism, while sharing practical results and future directions.
This article details the comprehensive optimization of 58's commercial traffic ranking system, covering data‑flow upgrades, advanced feature engineering, model enhancements—including online training, multi‑task and relevance models—and a multi‑factor ranking mechanism that together improve monetization efficiency and user experience.
This article details how Vimur, a Kafka‑based real‑time data pipeline, addresses the challenges of service splitting, database sharding, data migration, and synchronization by using CDC, a unified Avro format, and a change distribution platform to support search indexing, cache refresh, and reactive architectures.
