tracer-cloud--opensre/docs/comparisons/overview.mdx

---
title: 'How Tracer fits in your stack'
sidebarTitle: 'Overview'
description: 'Where Tracer fits in modern scientific and data platforms'
---

Modern scientific and data platforms are built from multiple layers: workflow orchestration, execution environments, infrastructure, and observability tooling. Each layer answers different questions, but gaps often appear between them, especially at runtime.

Tracer is designed to sit between orchestration and infrastructure, observing what actually executes on the system and mapping that behavior back to pipelines, runs, tasks, and tools.

This page explains where Tracer fits, what it adds, and how it complements the tools already in your stack.

## The typical stack (and where gaps appear)

Most bioinformatics, data, and HPC environments include some combination of:

- **Workflow orchestration**: Tools such as Seqera, Nextflow, Prefect, Dagster, Airflow, Flyte, or Slurm define what should run and when
- **Execution environments**: Containers, batch systems, Kubernetes, cloud instances, or HPC clusters execute the work
- **Observability and monitoring**: Tools such as Grafana, Prometheus, Datadog, or AWS CloudWatch collect and visualize reported metrics, logs, and traces

Each layer is effective within its scope, but none are designed to fully explain how execution actually behaves at runtime, especially for short-lived, heterogeneous scientific workloads.

## Where Tracer fits

Tracer observes execution directly from the host and container runtime. It does not replace orchestration or monitoring tools. Instead, it adds a missing layer:

**Execution-level visibility, grounded in what the operating system actually runs.**

Tracer answers questions such as:

- What is each pipeline step doing while it runs?
- Which tools are CPU-bound, I/O-bound, stalled, or idle?
- Which runs, tasks, or tools consumed resources and cost?
- Which infrastructure is active, idle, or orphaned after execution completes?

This visibility is derived from observed system behavior, not from reported metrics, labels, or manual instrumentation.

## How Tracer complements workflow orchestration

Workflow engines define pipeline structure, scheduling, retries, and state. They do not observe low-level execution behavior inside containers or processes.

Tracer complements orchestration tools by:

- Observing execution inside containers and hosts
- Capturing short-lived processes and subprocesses
- Mapping runtime behavior back to pipeline runs and tasks
- Providing execution context without modifying workflows

<Card title="Workflow orchestration tools" icon="diagram-project" href="/comparisons/workflow-orchestration">
  See how Tracer works with Airflow, Dagster, Flyte, Prefect, and Seqera
</Card>

## How Tracer complements observability and monitoring tools

Observability platforms collect telemetry that systems and applications report. They organize data around hosts, services, and metrics.

Tracer complements these tools by:

- Observing execution behavior directly, not via exporters
- Organizing data around pipelines, runs, tasks, and tools
- Providing cost and performance attribution at execution-unit granularity
- Surfacing behavior that occurs between metric scrapes or outside service boundaries

<Card title="Observability and telemetry tools" icon="chart-line" href="/comparisons/observability-tools">
  See how Tracer works with Datadog, Grafana, and Prometheus
</Card>

## How Tracer complements cloud-native monitoring

Cloud-native monitoring services such as AWS CloudWatch collect metrics, logs, and events from managed infrastructure. They are tightly integrated with cloud platforms but remain resource-centric rather than execution-aware.

Tracer complements these tools by:

- Observing execution behavior inside cloud workloads
- Mapping resource consumption to pipelines, tasks, and tools
- Providing visibility into short-lived and containerized execution
- Attributing cost to actual work, not just instance uptime

<Card title="Cloud-native monitoring" icon="cloud" href="/comparisons/cloud-native-monitoring">
  See how Tracer works with AWS CloudWatch
</Card>

## What Tracer does not try to be

Tracer is intentionally scoped. It does not aim to replace:

- Workflow orchestration engines
- General-purpose metric storage systems
- Organization-wide dashboarding for unrelated services
- Arbitrary business or application analytics

Tracer focuses on execution behavior, not on replacing every tool upstream or downstream.

## A shared mental model

A useful way to think about the stack:

| Layer | Role |
|-------|------|
| **Orchestration tools** | Define intent |
| **Infrastructure** | Execute work |
| **Monitoring tools** | Report signals |
| **Tracer** | Observe reality |

Tracer bridges intent and reality by showing how execution actually unfolds on the system.

## Where to go next

Choose the page that matches the tools in your environment:

<CardGroup cols={2}>
  <Card title="Workflow orchestration tools" icon="diagram-project" href="/comparisons/workflow-orchestration">
    Airflow, Dagster, Flyte, Prefect, Seqera
  </Card>
  <Card title="Observability and telemetry tools" icon="chart-line" href="/comparisons/observability-tools">
    Datadog, Grafana, Prometheus
  </Card>
  <Card title="Cloud-native monitoring" icon="cloud" href="/comparisons/cloud-native-monitoring">
    AWS CloudWatch
  </Card>
</CardGroup>

Each page explains what that tool does well, where execution-level visibility becomes important, and how Tracer integrates without changing existing workflows.

<div style={{ height: '50vh' }}></div>