Scaling Yandex search throughput by 4×
summary
Describes how Yandex increased search throughput fourfold by adapting to hardware loss and supply constraints; they analyzed bottlenecks and changed system architecture and resource usage to handle higher load with less hardware. Covers concrete engineering measures for scaling search capacity under constrained infrastructure (performance tuning, deployment and cluster adjustments).
AI-written summary · about
chapters (9)
- 0:00Introduction
- 3:59Datacenter shutdown problem
- 8:21Adaptive ranking methods
- 13:43Cache and load balancing
- 21:39Service acceleration and isolation
- 28:21Infrastructure changes
- 31:43Traffic goals 150% and 200%
- 36:40Degradation models and distillation
- 43:15Summary and Q&A
related talks
45mOptimizations for Yandex Search Results
39mOn the Present and Future of the Browser
29mSpeeding Up Image Preview Rendering in Yandex.Disk
41mScaling Search and Powering Vector-Based Retrieval With OpenSea...
19mEnhance Search Performance At Scale with
29mSpeed across all services
31mMaximize Resource Efficiency with Separated Index and Search Workloads
40mIn Search of... A better, fast