The article examines common pitfalls of naïve ad insertion in short‑video feeds, explains how cursor‑based pagination prevents duplicate ads, and outlines a client‑side mixing architecture that isolates services, meets strict latency requirements, and ensures accurate ad billing.
The paper introduces LEADRE, a multi‑faceted knowledge‑enhanced large language model‑driven display advertisement recommender that tackles user interest modeling, knowledge alignment, and low‑latency deployment, achieving significant GMV gains in Tencent’s ad platforms through innovative prompt engineering, semantic alignment, and TensorRT‑accelerated inference.
Bilibili's R&D built a cloud-based broadcast console for the 2024 CCTV Spring Festival Gala, delivering 4K HDR streaming, AI SDR-to-HDR conversion, low latency, bandwidth‑aware transcoding, and a synchronized “send bullet screen” interactive feature using custom SEI timestamps for hundreds of millions of viewers.
Async‑fork shifts the costly page‑table copying from Redis’s parent process to its child, allowing the parent to resume handling queries instantly and cutting snapshot‑induced latency spikes by over 98%, thereby dramatically improving tail latency during AOF rewrites, RDB backups, and master‑slave synchronizations.
This comprehensive guide explains Linux kernel performance profiling—both on‑CPU and off‑CPU—by stressing the need to target the critical 3 % of code, covering throughput, latency and power metrics, scalability laws, flame‑graph visualizations, perf and eBPF tools, lock‑contention analysis, and further reading recommendations.
This talk recounts how the Good Doctor platform tackled severe online incidents by launching the DOA project, then a service risk governance initiative that identifies, quantifies, and mitigates latency‑related risks through metrics‑driven development, dependency analysis, middleware reliability, and a dedicated risk‑management platform.
By replacing the default Parallel GC with a ParNew‑CMS collector, enlarging the Young generation, fixing Metaspace settings, and tuning CMS occupancy thresholds, the video service cut Young and Full GC pauses dramatically, lowered Full GC count by over 80%, and achieved more than 30% P99 latency reduction, with some APIs improving up to 80%.
By adding model warm‑up files, separating load/unload threads, switching to the Jemalloc allocator, and isolating TensorFlow’s parameter memory from RPC request buffers, iQIYI’s engineers reduced TensorFlow Serving hot‑update latency spikes in high‑throughput CTR recommendation services from over 120 ms to about 2 ms, eliminating jitter.
