Design of the aviasales Metasearch Engine
summary
The talk describes AviaSales’ metasearch engine for flight tickets and the custom processing engine called “Yasen” built to query and handle responses from dozens of external APIs. It explains handling large XML/JSON payloads (200–2000 KB per provider), and the design decisions to meet requirements: high availability, easy scalability, hot deploys, simple configuration, and low developer onboarding cost.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 3:19Old system problems
- 6:31Unit functional architecture
- 10:47Unit chain types
- 13:58DSL and system benefits
- 15:44Database organization
- 21:28Implementation details
- 22:33Cluster organization
- 26:25Tools and interfaces
- 30:22Results and advantages
- 31:30Q&A
related talks
41mThe evolution of search at Avito
47mWhere Your Ads Live
47mWhere Your Ads Live
34mMake It Under 100 Milliseconds: Contextual Ads on Sphinx
21mPreparing Yandex search data — the library and processes we built
56mLarge scale websites performance optimisation tricks
37mAvito.AutoUpload: From 4M to 80M Active Listings
45mOptimizations for Yandex Search Results
