Innovations in AWS analytics: Data processing
summary
EMR shows ~3.9x speed for Spark and ~2.7x for Trino out of the box; moving to Graviton 4 (transcript cites Graviton 3 → ~20% faster at 15% lower cost) yields further gains. Redshift metadata refresh is up to 146x faster, Redshift fleets example runs 283 clusters over 240 PB, and serverless reports up to 5.3x performance with ~70% cost reduction. Demos show zero ETL from DynamoDB and Salesforce into Redshift, Glue usage profiles to cap DPUs, MWAA faster executor scale‑down, and GenAI tools that troubleshoot Spark failures and auto-upgrade Glue/Spark jobs.
AI-written summary · about
chapters (8)
- 0:00Introduction
- 1:05Best price performance at lowest cost
- 5:58Redshift SQL engine innovations
- 13:06Operational excellence in analytics
- 17:55Zero ETL and data integration
- 19:19Live demo: Zero ETL with DynamoDB and Salesforce
- 23:48Generative AI for data processing
- 27:45Q&A and session wrap-up
related talks
01h 02mData warehousing reinvented for today’s needs
38mUnleashing the serverless data pipelines
43mFuture-proofing your applications with AWS databases
28mModernize Your Data Warehouse
22mHow Rocket Companies modernized their data strategy on AWS
30mSpeed Up Your Spark Data Processing with a GPU
22mTurn Data Into Insights
27mVisualizing Cloud Systems