Engineering the Cloud for Tomorrow: Performance, Speed and Scale for AI Workloads
summary
Oracle announces GA of 3‑rack dedicated regions, an OCI Dynamic Resource VM that reclaims roughly 50% idle memory at the kernel level, and liquid‑to‑air cooling that supports up to 1 MW per GPU rack. MI355X and MI300X GPU instances are available, NVIDIA HGX B300 and GB300 NVL72 appear for large jobs, and Zettascale10 stitches clusters with up to 800 GPUs delivering ~2.5× front‑end throughput and ~131 PB/s fabric bandwidth via RDMA and Accelleron DPUs.
AI-written summary · about
chapters (13)
- 0:00Oracle’s Vision: Cloud & AI for the Enterprise
- 2:32AI Changes Everything: The Challenges Ahead
- 5:07OCI Distributed Cloud: Services, Scale, and Global Reach
- 10:37Engineered for Flexibility, Performance, and Choice
- 14:31Compute, Network, and Storage: Innovations & Economic Benefits
- 19:59Memory, Price Efficiency, and Dynamic Resource VMs
- 23:39Uber’s Journey: Reliability, Velocity, Cloud Migration & AI Scale
- 29:38The Technical Path to Fast, Reliable Global AI Platforms
- 35:00Data Center Innovations: Liquid Cooling, Power Mgmt, & GPU Readiness
- 40:21Expanding Performance Choices: AMD, NVIDIA, Superclusters
- 49:14Partner Insights: AMD, NVIDIA, Arista Integration & OCI Networking
- 58:25Scale, Networking, Zettascale10, and the Future of AI Workloads
- 1:11:10Closing: OCI Performance, Accelleron, Speed, and Call to Action
related talks
11mthe future of cloud
23mScaling Up Your Cloud with Oracle Superclusters
18mClay Magouyrk on AI, Flexibility, Speed and Security with OCI
24mDevelopment Made Easy in the Cloud
01h 19mAccelerating Growth in the Cloud
01h 18mCloud Transformation Roadmap on Oracle Cloud Infrastructure
18mNext Gen Data Architecture for Multicloud
11mNVIDIA and the AI-Powered Enterprise