Monitor GPU Utilization and VRAM with DCGM Exporter

Problem: how not to lose money on GPUs?

Picture this: you start fine-tuning an LLM on 8 A100s. After 6 hours, OOM strikes—2000 iterations lost. Or your GPUs are at 2% utilization while you pay full price. Without monitoring, it's a gamble. We've seen it on 30+ projects: proper GPU monitoring with DCGM Exporter, Prometheus, and Grafana cuts downtime by 70% and saves up to 40% on infra costs. Control VRAM, utilization, and Tensor Cores—key for stable workflow.

We offer a turnkey GPU monitoring setup in 2–3 days. Contact us for a free assessment.

Why DCGM Exporter is the best choice for NVIDIA GPUs?

DCGM (Data Center GPU Manager) is an official NVIDIA tool. Unlike nvidia-smi, it exposes profile metrics like Tensor Cores, NVLink, and accurate SM load. Comparison:

DCGM gives 10x more metrics and natively exports to Prometheus, as confirmed by NVIDIA documentation.

Which metrics are critical for AI workloads?

For AI engineers, not only basic utilization and VRAM matter. Tensor Cores are key for training and inference performance. The metric DCGM_FI_PROF_PIPE_TENSOR_ACTIVE shows how efficiently these blocks are used. If it's low while GPU load is high, you're likely hitting memory or bus bottlenecks. Also critical is NVLink usage—in distributed training, the bottleneck is often interconnects.

Typical problems and solutions

How we set up monitoring turnkey

Step 1. Deploy DCGM Exporter

We install DCGM Exporter via Docker on each GPU node:

docker run -d --gpus all --cap-add SYS_ADMIN -p 9400:9400 --name dcgm-exporter nvcr.io/nvidia/k8s/dcgm-exporter:3.3.5-3.4.0-ubuntu22.04

For clusters, we use docker-compose with additional collector settings. We match the version to your OS and NVIDIA drivers.

Step 2. Configure Prometheus

We create a scrape target for DCGM metrics and alerting rules:

# prometheus.yml
global:
  scrape_interval: 15s
scrape_configs:
  - job_name: dcgm
    static_configs:
      - targets:
          - gpu-server-1:9400
          - gpu-server-2:9400
rule_files:
  - "gpu_alerts.yml"

Step 3. Set up alerts

We define key alerts: OOM (VRAM >95%), overheating (>85°C), low utilization (<20%), and service unavailability.

# gpu_alerts.yml
groups:
  - name: gpu_alerts
    rules:
      - alert: GPUMemoryNearFull
        expr: (DCGM_FI_DEV_FB_USED / DCGM_FI_DEV_FB_TOTAL) > 0.95
        for: 5m
        labels:
          severity: warning
        annotations:
          summary: "GPU {{ $labels.gpu }} on {{ $labels.instance }}: VRAM > 95%"
      - alert: GPUUtilizationLow
        expr: avg_over_time(DCGM_FI_DEV_GPU_UTIL[30m]) < 20
        for: 1h
        labels:
          severity: info
        annotations:
          summary: "Low GPU utilization on {{ $labels.instance }}"

How to set up an alert for OOM due to VRAM?

Add a rule that checks if VRAM exceeds 95% for 5 minutes. On trigger, you'll get a notification via Telegram or Slack. Additionally, you can set an alert on memory growth rate: if it increases by 10% in 2 minutes, it's a sign of impending OOM. We include such rules in the base configuration.

Step 4. Grafana dashboard

We build panels for utilization, VRAM, Tensor Cores, and temperature. Example VRAM panel:

{
  "title": "VRAM Usage %",
  "type": "gauge",
  "targets": [{
    "expr": "DCGM_FI_DEV_FB_USED / DCGM_FI_DEV_FB_TOTAL * 100",
    "legendFormat": "{{instance}} GPU{{gpu}}"
  }],
  "fieldConfig": {
    "thresholds": {
      "steps": [
        {"color": "green", "value": 0},
        {"color": "yellow", "value": 80},
        {"color": "red", "value": 95}
      ]
    }
  }
}

We create separate dashboards for training (LLM, CV) and inference, adding Tensor Core utilization, NVLink throughput, and p99 latency for inference. All dashboards are adapted to your models.

Step 5. Documentation and training

After deployment, we hand over: operation manual, alert scheme, dashboard access, and a short training for DevOps/ML engineers.

What's included

• Installation and configuration of DCGM Exporter on each GPU node
• Prometheus and alerting rules setup (Telegram/Slack)
• Grafana dashboards tailored to your tasks (training, inference)
• Integration with existing monitoring stack
• Documentation and team training
• 1 month of support after launch

Our experience and guarantees

Years of experience in MLOps. 30+ GPU infrastructure projects for AI startups and enterprises. Certified NVIDIA engineers. We guarantee 24/7 monitoring stability. Reach out for a tailored setup—we'll evaluate your infrastructure and propose the best solution.

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.