---
title: "Metrics"
description: "Track and analyze comprehensive pipeline metrics with Tracer"
---

## Overview

Tracer automatically captures detailed metrics for every process in your pipeline, providing deep visibility into resource usage, performance, and execution patterns without requiring any code changes.

## Key Metrics Tracked
### Resource Utilization

- CPU Usage: Per-process utilization, saturation levels, and hotspots
- Memory Usage: Allocation, peak consumption, and memory efficiency
- I/O Ops: Disk reads/writes, throughput, and bottleneck detection
- Network Activity: Bandwidth usage, transfer volume, and congestion
- GPU Utilization: Compute load and VRAM usage (when applicable)

<video
  autoPlay
  loop
  muted
  playsInline
  style={{ width: '100%', maxWidth: '100%' }}
>
  <source src="/images/Metrics.mp4" type="video/mp4" />
  Your browser does not support the video tag.
</video>

### Performance Metrics

- Execution Time: Wall-clock, CPU, and wait times per task
- Throughput: Data processed per unit time (e.g., samples/hour)
- Latency: Startup delays, scheduling lag, and queueing overhead
- Parallelization: Concurrency levels and contention across workers

### Cost Metrics

- Compute Costs: Per-task and per-pipeline cost attribution
- Resource Costs: Storage, network, and transfer charges
- Efficiency: Cost per sample, cost per GB processed
- Waste Detection: Idle resources, over-provisioning, and inefficiencies

<video
  autoPlay
  loop
  muted
  playsInline
  style={{ width: '100%', maxWidth: '100%' }}
>
  <source src="/images/Costs.mp4" type="video/mp4" />
  Your browser does not support the video tag.
</video>

## Real-Time Monitoring

Inspect live metrics as your pipeline runs, with immediate visibility into performance, utilization, and potential bottlenecks.

## Historical Analysis

- Analyze past runs to uncover long-term patterns and optimization opportunities:
- Performance Trends: How execution times shift across versions
- Utilization Patterns: Persistent over/under-use of compute or memory
- Cost Evolution: Impact of configuration changes on total spend
- Regression Detection: Automatic alerts for performance degradation