Kubernetes for AI/ML: GPU Scheduling with NVIDIA GPU Operator

In a production cluster with 8× NVIDIA A100, we faced GPU utilization of only 40% due to lack of MIG and improper scheduling. After implementing the NVIDIA GPU Operator and gang scheduling, utilization rose to 92%, and idle time dropped by 55%. Our team has 5+ years of experience configuring GPU clusters for AI/ML and has delivered over 20 projects. Without proper GPU scheduling, you risk losing up to 60% of resources — direct losses. Our experience shows that correct Kubernetes configuration for AI/ML workloads guarantees GPU utilization above 90% and reduces GPU infrastructure costs by 30–50%. Have your cluster audited — we'll identify bottlenecks and suggest optimization.

How the NVIDIA GPU Operator Simplifies GPU Management

The operator automatically deploys NVIDIA drivers, container toolkit, and device plugin — all via Custom Resources. Manual configuration on each node is eliminated. The result is a cluster with GPU support ready in one Helm release, 3 times faster than manual setup. The NVIDIA GPU Operator documentation confirms this.

helm repo add nvidia https://helm.ngc.nvidia.com/nvidia
helm repo update
helm install gpu-operator nvidia/gpu-operator \
  --namespace gpu-operator --create-namespace \
  --set driver.enabled=true --set driver.version="545.23.06" \
  --set toolkit.enabled=true --set devicePlugin.enabled=true \
  --set dcgmExporter.enabled=true --set gfd.enabled=true
kubectl get pods -n gpu-operator
kubectl get nodes -o custom-columns='NAME:.metadata.name,GPU:.status.capacity.nvidia\.com/gpu'

Why Proper GPU Scheduling Matters

Typical issues: a job cannot start because whole GPUs are unavailable, even though free MIG slices exist; or one task blocks the entire node. The solution is a combination of MIG, gang scheduling, and priorities.

The standard Kubernetes scheduler does not support gang scheduling, causing deadlock in distributed training. Volcano and Kueue solve this: Volcano is mature with fair sharing; Kueue is a native API. For large clusters, Volcano delivers 85–95% utilization versus 60–75% for the standard scheduler — a 30% improvement.

How to Avoid Deadlock in Distributed Training

Gang scheduling with Volcano ensures all pods of a distributed task start simultaneously:

apiVersion: batch.volcano.sh/v1alpha1
kind: Job
metadata:
  name: distributed-training
spec:
  minAvailable: 4
  schedulerName: volcano
  tasks:
    - replicas: 1
      name: master
      template:
        spec:
          containers:
            - name: master
              image: your-registry/pytorch-trainer:v1
              resources:
                limits:
                  nvidia.com/gpu: 8
    - replicas: 3
      name: worker
      template:
        spec:
          containers:
            - name: worker
              image: your-registry/pytorch-trainer:v1
              resources:
                limits:
                  nvidia.com/gpu: 8

Why Use MIG?

MIG (Multi-Instance GPU) on A100/H100 splits a GPU into up to 7 instances. Switching to MIG increases task placement density and reduces GPU costs. Request in a pod spec:

resources:
  limits:
    nvidia.com/mig-1g.10gb: 1

How to Set Priorities for ML Tasks

For production inference, set a PriorityClass with high priority (1000); for batch training, low priority (100) with PreemptLowerPriority. Critical services always get GPUs first; training runs on leftover resources.

What's Included in Cluster Setup

• Deployment of the NVIDIA GPU Operator with drivers and DCGM Exporter.
• PriorityClass for inference prioritization.
• Node labels for grouping GPU types (A100, V100, A10G).
• Gang scheduling (Volcano or Kueue).
• Cluster Autoscaler for dynamic GPU node scaling.
• Grafana dashboard with utilization, temperature, and NVLink metrics.
• Documentation and team training (optional).

How We Do It: Phases

1. Infrastructure audit — measure p99 latency, utilization, identify bottlenecks.
2. Design — choose a scheduler, define resource profiles and MIG configurations.
3. Implementation — Helm releases, monitoring setup, tests with synthetic loads.
4. Testing — verify FLOPS, low-priority job preemption, gang scheduling correctness.
5. Deployment and support — hand over documentation, activate monitoring, provide SLA.

Common mistakes when configuring a GPU cluster:

• Missing node labels — pods don't know which GPU type is on the node. Solution: kubectl label node gpu-node-1 nvidia.com/gpu.product=A100-SXM4-80GB.
• Unconfigured Cluster Autoscaler — the cluster doesn't grow under peak load. Specify min/max GPU nodes.
• No monitoring — performance degradation goes unnoticed. DCGM Exporter + Grafana provide full visibility.

Comparison of GPU Scheduling Approaches

Proper GPU scheduling in Kubernetes is key to efficient AI/ML workloads. We guarantee GPU utilization above 85% and stable operation even under peak load. Contact us for a cluster audit — we'll assess current utilization and suggest optimization. Get a consultation and a detailed implementation plan.

MLOps: Infrastructure for Training, Deploying, and Monitoring ML Models

The model is trained, metrics — F1 0.94 on validation. Three months later in production, quality drops by 12%. No one knows when — there is no monitoring. It's impossible to retrain quickly — the training script is in a Jupyter notebook of a data scientist who has already left. Data for retraining is collected manually from three disparate systems. About half of the projects come to us with this pain. We build a turnkey MLOps platform: from experiment tracking to automatic deployment and data drift monitoring. We will assess your infrastructure in 1–2 weeks, and in 4–6 weeks you will get a basic MLOps core running in production. Our team has 10+ years of experience in ML infrastructure, over 50 implementations.

How does MLOps infrastructure benefit your ML projects?

Experiment Tracking and Reproducibility

Without tracking, an ML project turns into chaos: it's unclear which checkpoint is better, which hyperparameters were used, which dataset. Reproducing a result a month later is a quest.

Why is experiment tracking the foundation of reproducibility?

MLflow is an open source standard for tracking. It logs parameters, metrics, artifacts (models, graphs), and code. MLflow Model Registry is a centralized model storage with versioning and lifecycle stages (Staging → Production → Archived). Deployment via MLflow Serving or integration with external systems.

Typical initialization in code:

import mlflow

mlflow.set_experiment("fraud-detection-v2")
with mlflow.start_run():
    mlflow.log_params({"learning_rate": 3e-4, "batch_size": 64, "epochs": 10})
    mlflow.log_metric("val_f1", val_f1, step=epoch)
    mlflow.pytorch.log_model(model, "model")

This is the minimum. In production, we add logging of system metrics (GPU utilization, memory), dataset (hash, version), code (git commit hash). Weights & Biases — richer UI, collaboration features, sweep for hyperparameter optimization. MLflow — for on-premise deployment without external dependencies.

DVC (Data Version Control) — versioning of data and models on top of git. Data is stored in S3/GCS/Azure Blob, only metadata (hashes) in git. dvc repro reproduces the entire pipeline from raw data to metrics.

To ensure reproducibility of training, fix random seeds (torch.manual_seed, numpy.random.seed, random.seed) and record them in experiment metadata. Without this, debugging irregular results is painful. Log the dataset version (DVC hash) and git commit — then any experiment can be reproduced down to the byte.

Pipeline Orchestration: Kubeflow, Airflow, Prefect

A pipeline orchestrator becomes necessary when: A 100-line training script in cron is fine for simple tasks. But as soon as you have a multi-step pipeline (data loading → preprocessing → feature engineering → training → validation → deployment if quality above threshold), you need an orchestrator with retry logic, visualization, and alerts.

Kubeflow — Kubernetes-native orchestrator for ML (see Kubeflow). Each step is a Docker container. Supports parallel steps, conditional branches, artifacts between steps. Integrates with Katib (AutoML), KServe (serving), Feast (feature store).

Apache Airflow — more general DAG orchestrator. Wide ecosystem of operators (S3, Spark, DBT, Kubernetes). Easier to deploy if Airflow already exists in the company.

Prefect / Metaflow — less boilerplate. Prefect 2.x with @flow and @task decorators — quick start for small teams.

Typical training pipeline architecture on Kubeflow:

1. Data ingestion component — fetches data from S3/DB, validates schema via Great Expectations
2. Preprocessing component — transformations, normalization, train/val/test split
3. Training component — training on GPU, logging to MLflow
4. Evaluation component — metric calculation, comparison with baseline in Model Registry
5. Conditional deployment — deploy only if new model is better than current by >2% F1

Each component is a separate Docker image. Pipeline is versioned in git. Scheduled run (retraining once a week on new data) or manual.

Model Registry and Lifecycle Management

Model Registry is not just a checkpoint store. It is a centralized system that knows:

• Which model is currently in production (and with what metrics)
• History of all versions with training parameters
• Metadata: dataset, git commit, validation results
• Lifecycle stage: None → Staging → Production → Archived

MLflow Model Registry — standard. For enterprise — Vertex AI Model Registry (GCP), SageMaker Model Registry (AWS), Azure ML Model Registry.

Model promotion through stages: automatically move model to Staging after successful eval, then manual or automatic (during A/B test) promotion to Production. Rollback — switch to previous Production version in seconds.

Serving: From FastAPI to Triton Inference Server

Simple case. FastAPI + PyTorch/ONNX on one server — 80% of production ML deployments are exactly that. Sufficient for most tasks with load up to 100 req/s.

from fastapi import FastAPI
import onnxruntime as ort

app = FastAPI()
session = ort.InferenceSession("model.onnx", providers=["CUDAExecutionProvider"])

@app.post("/predict")
async def predict(request: PredictRequest):
    inputs = preprocess(request.text)
    outputs = session.run(None, {"input_ids": inputs})
    return {"label": postprocess(outputs)}

Triton Inference Server — production standard for high loads (500+ req/s). Dynamic batching, concurrent model execution, model ensemble. Supports TensorRT, ONNX, PyTorch TorchScript, TensorFlow SavedModel.

KServe — Kubernetes-native ML serving with autoscaling, canary deployments, A/B testing out of the box. Scale-to-zero for inactive models — savings on infrastructure up to 40% annually for a project with 10 models.

Monitoring: Data Drift, Model Drift, Infrastructure Metrics

Monitoring — what is usually done last and regretted first. Three levels.

Infrastructure monitoring. Latency (P50/P95/P99), throughput (req/s), error rate (4xx, 5xx), GPU/CPU utilization. Prometheus + Grafana — standard. Alert when P99 latency > threshold or error rate > 1%.

Data drift monitoring. Distribution of input data changes over time. Detect via PSI (Population Stability Index) for numerical features: PSI > 0.2 — strong drift. Chi-squared test for categorical, Kolmogorov-Smirnov test for continuous. Evidently AI — open source library with ready-made drift tests.

Model drift monitoring. If ground truth is delayed (e.g., we know conversion after a week) — monitor real metrics. If not — surrogate metrics: distribution of prediction scores, proportion of confident predictions.

Alerting. Three levels: INFO (minor drift, log it), WARNING (significant, notify team), CRITICAL (quality dropped below threshold — automatic switch to fallback model).

Why is data drift monitoring important?

Without it, you learn about model degradation only from user complaints or ringing SLA. A drift alert allows you to retrain the model in advance, before errors start causing losses. In one of our projects, PSI monitoring detected drift 2 days after a data source change — this saved the campaign.

Feature Store

Feature Store solves the training-serving skew problem. If preprocessing during training and inference is implemented in two different places — divergence is inevitable.

A Feature Store is needed when:

• Several models use the same features
• Features are computed from streaming data (real-time)
• Large team with different people on feature engineering and model training

Feast — open source Feature Store. Offline store (S3 + Parquet) for training, online store (Redis, DynamoDB) for low-latency inference. Feature definitions as code, materialization job syncs offline → online.

Tecton (commercial), Vertex AI Feature Store (GCP), SageMaker Feature Store (AWS) — managed options with less ops overhead.

CI/CD for ML

ML CI/CD is regular CI/CD plus specific ML steps.

ML-specific checks in CI:

• Reproducibility check: run training with a fixed seed, result must match
• Data validation: Great Expectations or Pandera on schema/distribution checks
• Model performance check: automatic eval on holdout, block merge if degradation > threshold
• Latency regression test: inference must meet SLA

GitOps for deployment. Merge to main → CI triggers training → eval → if passes → automatic deployment to Staging → smoke tests → manual promotion to Production or automatic upon successful canary.

Tools: GitHub Actions / GitLab CI for CI, ArgoCD for GitOps deployment on Kubernetes.

What's Included in MLOps Platform Development

We provide a full cycle of work, documentation, and team training.

Total time from audit to full MLOps platform: 3–5 months. Also possible phased launch: basic level (tracking + serving) in 4–6 weeks.

Cost is calculated individually based on data volume, number of models, and infrastructure requirements. Order an MLOps infrastructure audit — get a roadmap in 1–2 weeks. Contact us for a project assessment — we will send a preliminary estimate within 2 business days.

Note: warranty on architectural solutions — 12 months. We provide integration certificates with major cloud providers (AWS, GCP, Azure). During our work, we have not lost a single client after the first implementation — the experience of 50+ successful MLOps projects speaks for itself. Get a consultation on building an MLOps platform today.