Ensuring 100% uptime in the cloud of failing GPUs
summary
Describes how Viz.ai runs performance-critical deep‑learning CT scan analysis on AWS GPU instances and uses dynamic allocation to manage cost and traffic. Recounts detecting a faulty GPU hardware on an AWS instance and explains the automated detection and mitigation mechanisms put in place to maintain continuous uptime for the service.
AI-written summary · about
related talks
29mThe GPUs on the Bus Go ‘Round and ‘Round
17mData centre security
15mMLaaS in the Wild: Workload Analysis and Scheduling in Large-Scale Heterogeneous GPU...
21mExperience Operating Large GPU Clusters at Organizational Scale
28mGPU sharing done right: Secrets, security, and scaling with Vault and vCluster
21mSponsored Session: Fault-Tolerant Training: How We Build Rel...
23mBuilding reliable AI hardware
17mTrustworthy, Sovereign AI: Creating Resilient Spaces