Architecture and Practices of the 58.com Alliance Advertising Platform
This article presents a comprehensive overview of the 58.com alliance advertising platform, detailing its business model, core modules such as SSP, direct‑investment, DSP and creative platforms, media integration flow, performance optimizations, GC tuning, indexing strategies, and future directions for programmatic creative generation.
The 58.com alliance advertising platform enables internal advertisers to acquire external traffic through SSP and DSP mechanisms, turning site traffic into revenue. Core modules include the Same‑City SSP for media partners, a direct‑investment platform built on marketing APIs, an internal DSP that connects to external ADX, and a programmatic creative platform for automated ad generation.
The media platform allows publishers to register companies, media, and ad slots, obtain placement links or JavaScript, and serve ads by invoking the alliance ad API, which routes requests to the internal ad inventory and reports performance data back to the analytics pipeline.
Effect evaluation employs three attribution methods—URL parameters, log tracing, and cookie‑based tracking—with cookie attribution being the preferred low‑maintenance approach.
Anti‑hijacking measures include enforcing HTTPS, custom DNS, and DNS‑over‑HTTPS to prevent DNS or router‑level redirection attacks.
DSP performance is critical; the system targets sub‑50 ms response times for real‑time bidding across ~100 k QPS. Optimizations include moving from a blocking thread model to an EventLoop‑Thread architecture with Future/Promise for asynchronous processing, reducing thread count and latency.
Garbage‑collection issues caused frequent Full GC pauses; a custom Free‑GC solution using a fixed‑size long array for city‑ID caching reduced GC frequency from every 2 hours to 30 hours and lowered TP99 latency from 80 ms to 50 ms.
Indexing for the DSP uses a dual‑shard design with a route‑key based on city and category, ensuring that 95 % of queries hit a single shard, and a route‑strategy stored in Zookeeper to simplify scaling and re‑sharding.
The programmatic creative system (Creative‑Builder) extracts key ad elements, selects appropriate templates, applies rule‑based optimization, renders images, uploads them to CDN, and indexes them for CTR‑based selection, closing the data loop with feedback from impressions and clicks.
Future work focuses on enhancing the creative generation pipeline, improving routing efficiency, and expanding the platform’s capabilities for broader advertising scenarios.
DataFunTalk
Dedicated to sharing and discussing big data and AI technology applications, aiming to empower a million data scientists. Regularly hosts live tech talks and curates articles on big data, recommendation/search algorithms, advertising algorithms, NLP, intelligent risk control, autonomous driving, and machine learning/deep learning.
How this landed with the community
Was this worth your time?
0 Comments
Thoughtful readers leave field notes, pushback, and hard-won operational detail here.