Best practices for advertising and marketing technology on AWS
summary
Cloud patterns and hard numbers: EKS with isolated Spark workers plus Lambda dispatches handle spiky ingestion and avoid queuing, yielding a reported 55% compute saving; real‑time bidding customers process ~800 billion transactions/day (≈30M/sec) with <20 ms latency. Try Graviton2 for 1.5–2× cost/perf gains, run EMR and data pipelines on Spot (some customers use ~80% Spot), convert text to ORC/columnar to cut EC2 by ~69%, and use SageMaker to shrink ML resource use ~60% and costs ~50% while speeding experiments from months to weeks.
AI-written summary · about
chapters (22)
- 0:00Intro
- 1:20Data center onprem deployment
- 2:18Demo
- 3:48Challenges
- 5:18Outcome
- 6:21Realtime advertising
- 7:08Graviton II
- 8:20Trade Desk Example
- 9:18Test Instance Types
- 10:30Migration
- 12:17Customer Data Identity Resolution
- 13:17Data Processing
- 13:38Format Matters
- 14:15Data ingestion
- 15:02Storage
- 16:49Identity load
- 17:30Innovation and speed
- 18:21Machine learning challenges
- 19:25Machine learning development environment
- 20:28Data selection and engineering
- 22:20Results
- 23:25Summary
related talks
19mFireside Chat
44mBest practices for analytics and generative AI on AWS(ANT302)
22mHow Rocket Companies modernized their data strategy on AWS
46mArchitecting for Success and Runway
34mBuilding scalable end-to-end deep learning pipelines in the cloud
32mOptimizing AI/ML workloads for sustainability
16mAWS On Air Voice of the Customer ft. Amazon Ads
40mReady, set, data: Perfecting the art of talent matching with AI