Amazon Redshift Reinvented
summary
Launched in 2013, Redshift claims tens of thousands of customers analyzing exabytes daily and now separates compute from storage with Redshift Managed Storage to scale to tens of petabytes per database. Query compilation generates tuned C++ and uses vectorized execution; performance improvements push price/performance 1.2–3.8× over peers and raise throughput from ~200 QPH to 12,000 queries per hour (≈60×). Features added include Auto tables for automatic physical design, Auto Scaling and cross-cluster attach, Spectrum, Federated Query, SageMaker in-SQL training/inference, and incremental materialized views.
AI-written summary · about
chapters (26)
- 0:00Data analytics challenges
- 1:53Target data problem
- 2:55Redshift evolution overview
- 3:38Security and availability
- 4:19Redshift architecture
- 6:27Performance improvements
- 7:29Price performance superiority
- 8:11Linear scalability
- 9:12More data and concurrency
- 9:53Disaggregated storage and compute
- 10:56Auto-scaling clusters
- 11:39Concurrency scaling achievements
- 13:06Data sharing
- 14:32Autonomics reduce operations
- 15:36Detecting gray errors
- 16:39Automated maintenance
- 17:40Auto WLM table management
- 19:07Auto WLM performance gains
- 20:09AWS ecosystem integration
- 20:52Redshift Spectrum
- 21:33Federated queries
- 22:17Incremental materialized views
- 22:58Machine learning integration with SageMaker
- 23:44ML in SQL queries
- 24:06Lambda custom functions
- 24:52Redshift evolution summary
related talks
42mHow innovation happens on Amazon Redshift
48mData warehousing reinvented for today’s needs
23mOne day with Redshift
22mRedshift Serverless - easily run big data analytics at scale
43mFuture-proofing your applications with AWS databases
22mReference architectures for Amazon Redshift data warehouses
28mInnovations in AWS analytics: Data processing
41mOwning Your Own (Data) Lake House