Leveraging Kubernetes for Machine Learning

https://www.youtube.com/watch?v=vlMaD5VY-co&t=17s

Transcript

1. Machine Learning with Kubernetes Christopher M Luciano - IBM
2. ./speakerQuery.sh • Christopher M Luciano • Advisory Software Engineer @ IBM • Part of Open Source Technology Team in IBM Digital Business Group • Contributor to Kubernetes (SIG Network & SIG Node) • Work on Cloud Native Computing Foundation (CNCF) projects • @cmluciano_ on Twitter • github.com/cmluciano
3. My blog
4. Machine Learning Part of the @cmluciano_ What Where Why Series
5. Error Correction
6. The Goal • Base Knowledge -> • Points of Analysis -> • Corpus of Unstructured -> • Error Correction -> • Rinse and Repeat
7. Sprinkles Distinct Features • Ears • Face • Feet • Tail
8. Sprinkles Features • Face • Ears • Feet
9. Penguin Cat? What do we know? Do we see features? Conclusion
10. Duck Short snout Circular face Eyes not right No fur
11. Onyx Features? • Ears • Face • Snout • Feet
12. Sloth Circular face Short nose Narrow eyes Distinct mouth Hair Feet
13. Cats? Face Ears Nose Tail Feet
14. Watson
16. Watson Developer Cloud
17. IBM X-Force https://www.ibm.com/security/xforce/
18. How Can I Do This Myself • GPUs • Kubernetes • Tensorflow
19. Stacks on Stacks on Stacks Bare-metal -> Openstack -> Virtual Machines -> Runtime -> Kubernetes -> Tensorflow
20. Think of an Irish Breakfast
21. GPU Characteristics • Multiple Video Cards • Driver installation • Heterogenous model distribution • Resource fragmentation • Failure scenarios
22. Multiple Video Cards http://www.softlayer.com/GPU
23. Driver Installation • Host - container driver sets • Kubernetes daemonsets
24. Heterogenous Video Cards Node selectors https://kubernetes.io/docs/tasks/manage-gpus/scheduling-gpus/#api
25. Resource Fragmentation • Scheduler priorities • Topology
26. GPU Failure Scenarios • Fan heat • Insufficient power • ...
27. Kubernetes 1.6 GPU features • Multi-GPU pod scheduling • Video card discovery • Basic failure recovery • Only works with Docker
28. Where Are We Going? • Advanced – Device Recovery – Health checking features – Topology – Metrics, metrics, metrics • Cleanup – Support in Container Resource Interface (CRI) • Container runtime independence – Use of NVML or libnvidia-container
29. Ask Me Anything on Kubernetes • Use Cases for GPU/HPC (High Performance Computing) • Kubernetes Networking • Kubernetes Features • Prometheus • Other CNCF
30. If You Just Want to Talk • Cars • Coffee • Cooking • Fishing • World Culture
31. Questions? • GPUs • Kubernetes • ….
32. Thank You! • Christopher M Luciano • Advisory Software Engineer @ IBM • Part of Open Source Technology Team in IBM Digital Business Group • Contributor to Kubernetes (SIG Network & SIG Node) • Work on Cloud Native Computing Foundation (CNCF) projects • @cmluciano_ on Twitter • github.com/cmluciano