the future of cloud
summary
Oracle outlines concrete infrastructure wins: a packaged OCI Supercluster that pairs Nvidia GPUs with RDMA-over-converged-Ethernet networking and scales to 16,000 GPUs, delivering up to 3.2 TB/s between racks and quoted inter‑GPU latencies near 20 microseconds, so training runs faster and costs less. Multi‑cloud work includes an Oracle Database at Azure interconnect used by AT&T, Vodafone and Voya. Most cloud services and Fusion workloads are already running on Ampere ARM silicon, with 95% migrated to Ampere for big energy and cost reductions.
AI-written summary · about
chapters (8)
- 0:00Introduction
- 1:10AI’s impact on cloud
- 2:58Multi-cloud flexibility
- 4:23OCI Supercluster for AI/ML
- 5:25RDMA over converged Ethernet
- 7:53OCI growth drivers
- 9:16Sustainability in cloud
- 10:44Closing remarks
related talks
23mScaling Up Your Cloud with Oracle Superclusters
01h 13mEngineering the Cloud for Tomorrow: Performance, Speed and Scale for AI Workloads
18mClay Magouyrk on AI, Flexibility, Speed and Security with OCI
01h 18mCloud Transformation Roadmap on Oracle Cloud Infrastructure
01h 19mAccelerating Growth in the Cloud
56mKeynote on Building the Cloud for You
24mDevelopment Made Easy in the Cloud
10mOracle's Approach to Multicloud