Workload Management

NCP-AIO · 46 questions

• City needs a 311 text-classifier inference service on Kubernetes. How should the workload be deployed?
• A public-health symptom triage model will run on Kubernetes. How should its network exposure be deployed?
• Transit ETA inference must survive a single depot GPU node failure. What Kubernetes deployment pattern fits?
• Library branch kiosks need continuous catalog-vision inference. Which Kubernetes workload type should ops deploy?
• Water leak-detection inference is getting a new model. How should the Kubernetes rollout start?
• Courts OCR online inference must land on a specific GPU product class. What should the Kubernetes deploy request?
• Emergency chat-assist inference is quiet off-shift. How should Kubernetes deployment save civic GPU budget?
• Housing authority needs a policy-Q&A model served with predictable GPUs. How should inference be deployed?
• 311 Run:ai inference must stay responsive during Monday morning ticket storms. What should ops configure?
• Parks computer-vision inference should consume only parks capacity. How should the Run:ai workload be deployed?
• County assessor needs a new imagery model version in production. How should inference roll out?
• School-district demo day apps need temporary Run:ai inference access. What deployment control fits?
• Transit platform policy must stay centralized for ETA inference serving. What deployment approach should ops choose?
• Public-health needs a multi-node fine-tune on the civic Slurm cluster. How should the training workload be deployed?
• Transit overnight ETA training must process each weekly dataset chunk. Which Slurm pattern should ops deploy?
• Assessor imagery training must avoid MIG-sharing interference. How should the Slurm workload request capacity?
• Water-utility hyperparameter sweeps need many related Slurm training jobs on shared data. What packaging fits?
• Courts OCR model training crawls when the Slurm working directory sits on slow NFS. What placement fix fits?
• Emergency-management needs a surge retrain during a city drill. How should the Slurm training job be deployed?
• A library consortium must train a shared embedding model across two campus GPU sites. What Slurm approach fits?
• City data science needs distributed 311 model training on the municipal Run:ai platform. How should the workload be deployed?
• A public-health Run:ai training job requests more GPUs than the project quota allows. What should operators expect?
• Housing-authority fine-tunes must shrink when inference priority rises. What Run:ai training pattern fits?
• School-district teachers need to launch GPU training from notebooks without reinventing cluster settings. What fits?
• Transit ops compare Slurm and Run:ai for the same ETA training job on a Kubernetes-centric platform. Which choice fits?
• Parks needs a seasonal computer-vision retrain on a predictable cadence. How should Run:ai training be scheduled?
• City NOC suspects ECC errors on a 311 GPU node before escalating a ticket. What first-line check fits?
• Public-health training slowdowns have no clear application bug. What system tool triage fits?
• DGX-class civic systems should warn about fan and PSU problems before models fail loudly. What tool fits?
• Transit sees Run:ai jobs pending while teams claim the code is broken. What correlation helps capacity decisions?
• Library kiosk GPUs appear slow and nvidia-smi reports thermal throttle. What should ops conclude first?
• County ops want weekly GPU health trends in the municipal observability stack. What integration fits?
• Water-utility ops need GPU diagnostics on one suspect node without taking the whole partition down. What fits?
• School-lab GPUs disappear from nvidia-smi after a driver change. What should ops do before reimaging?
• After budget hearings, the city must split GPUs across 311, health, and transit on Run:ai. What allocation fits?
• Housing under-uses Slurm GPUs while parks over-uses. How should civic team allocation be adjusted?
• A student namespace on the school-district cluster risks claiming all GPUs. What Kubernetes control fits?
• An emergency drill needs temporary GPUs moved from libraries to emergency-management pools. What fits?
• Assessor and clerk teams keep over-promising GPUs. What cross-team allocation practice fits?
• 311 Run:ai burst must not starve health training for an entire quarter. What allocation balance fits?
• A mixed Slurm and Kubernetes civic site must allocate nodes to teams without double-booking. What practice fits?
• After a CVE patch, 311 serving must pick up an updated NGC TensorRT container at runtime. What fits?
• Public-health training must start from an NGC PyTorch container matched to the cluster CUDA level. What should ops pick?
• Transit wants GPU workload base images that stay supportable and reject unsigned sidecars. What practice fits?
• City must expand 311 capacity mid-quarter by adding a new GPU node to the existing BCM cluster. What fits?
• After hardware delivery, county must finish BCM provisioning so the node joins the right category and scheduler. What should ops do?