Qunar Tech Salon is a learning and exchange platform for Qunar engineers and industry peers. We share cutting-edge technology trends and topics, providing a free platform for mid-to-senior technical professionals to exchange and learn.
This article details the background, challenges, migration plan, automation, and performance‑tuning techniques used to relocate a petabyte‑scale Elasticsearch logging cluster from one data center to another, highlighting practical lessons and measurable improvements in stability and migration speed.
This article describes Qunar's use of traffic recording and replay technology for fault isolation, performance optimization, and upgrade migration, detailing the evolution of their Java‑based agents, the end‑to‑end testing workflow, full‑link stress‑test integration, and the measurable business and performance outcomes.
This article describes the background, technology selection, architecture design, data flow, monitoring, logging, and trace collection mechanisms of Qunar's self‑built distributed tracing system, analyzes major performance problems such as Flume interruptions, Kafka bottlenecks, Flink back‑pressure, and presents concrete solutions including sliding‑window throttling, CGroup limits, and JavaAgent instrumentation, ultimately improving trace connectivity and system observability.
This article describes the challenges of migrating a long‑running monolithic database to a sharded architecture, presents a two‑stage smooth migration strategy using double‑write, diff verification, and mapping‑key techniques, and details the componentized design, core implementation, and configuration of a reusable sharding framework built on MyBatis and Spring.
In this interview, Qunar VP Sun Bin reflects on his 13‑year journey, the technology operations center’s pandemic‑driven innovations, the company’s AI committee and big‑data strategies, and the philosophy of pure technical branding within the ITCP alliance.
This article presents a detailed case study on how Qunar's front‑end team reduced the Time to Interactive of the hotel‑detail page by introducing pre‑fetching, cache‑key redesign, view‑model batching and longer cache lifetimes, achieving up to a 70% TTI improvement on Android devices.
The QTest and QClient technical journey event on May 23, 2023 features multiple expert talks on performance testing, AI‑enhanced frontend engineering, Serverless visualisation, user‑behavior replay, and a new React framework, along with registration details, promotional items, and recruitment information.
This article presents Qunar's comprehensive system slimming initiative, detailing the background of growing code bloat, the challenges of reducing millions of lines across thousands of services, the two‑phase plan for service and code pruning, the use of SA and other tooling to identify and safely delete unused resources, and the measurable outcomes such as nearly 50% code reduction, improved development speed, and future directions for further optimization.
The article details the background, architecture, core features, alert governance, trace integration, and cloud‑native evolution of Watcher, Qunar's internally built, highly scalable monitoring platform that unifies application‑level metrics, alerting, and observability across thousands of services and containers.
This article describes the design, implementation, and evolution of Heimdall, an exception‑statistics platform built on Kafka, Flink, and HBase that provides minute‑level anomaly aggregation, stack trace querying, and integration with release and alerting workflows to improve service reliability across thousands of micro‑services.
