Skip to main content
Tracer is an execution-level observability platform designed for scientific and compute-intensive workloads. This page describes the capabilities, compatibility, and guarantees that apply across all Tracer products. For details on how execution data is collected, analyzed, or acted on, see:

Tracer/collect

Tracer/tune

Tracer/sweep

​
Execution-level observability

Tracer observes workloads at the operating system level. It captures how processes execute and use resources, rather than relying on workflow metadata, logs, or application instrumentation.

Real behavior

Observes real execution behavior, not configuration intent

Zero changes

Works without modifying code, containers, or workflows

No tagging

Does not require tagging or framework-specific integration
This execution-first approach is consistent whether Tracer is used for pipeline analysis, optimization, or systemwide cost discovery.

​
Supported environments

Tracer is designed to work in real production environments with heterogeneous tooling.

​
Operating system

  • Linux hosts only

​
Infrastructure

  • Cloud compute (for example, AWS EC2 and AWS Batch)
  • On-premises Linux clusters
  • Hybrid cloud and on-prem environments

​
Workload types

  • Containerized and non-containerized workloads
  • Batch processing systems
  • Interactive compute sessions
  • Pipelines with many short-lived processes
  • Legacy tools and custom binaries
If a workload runs on Linux, Tracer can observe it.

​
Framework and workflow support

Tracer is framework-agnostic. It does not depend on workflow engine metadata and does not require configuration for specific systems. Commonly used with:
  • Nextflow
  • Snakemake
  • CWL
  • Slurm
  • Custom scripts and pipelines
  • AWS Batch–managed workloads
This list is illustrative, not exhaustive.

​
Installation guarantees

Across all supported environments:

No code changes

Works with existing code as-is

No instrumentation

No application wrappers required

No tagging

No annotation or tagging required
Once Tracer is installed, execution visibility is available automatically on the next run. See Quickstart for setup instructions.

​
Security and data boundaries

Tracer is designed to minimize data exposure.
Across all products, Tracer:
  • Collects system-level execution metadata only
  • Does not inspect application payloads or scientific data
  • Does not read file contents or memory
  • Does not collect environment variables or secrets
Detailed data-collection limits and privacy boundaries are documented in Limits and privacy. Information about kernel-level safety and isolation is covered in eBPF and security.

​
Performance characteristics

Tracer is built for continuous use in compute-heavy cloud environments. (Also works locally) General characteristics:
  • Low overhead — Less than 2% runtime overhead
  • No re-runs — No need to re-run pipelines to collect data
  • Scales well — Handles thousands of short-lived processes
Specific performance details depend on workload characteristics and are documented in Tracer/collect.

​
What this page is for

This page answers:
  • Will Tracer work in my environment?
  • What are the system requirements and guarantees?
  • What assumptions does Tracer make about workflows and infrastructure?
It intentionally does not describe product-specific behavior or UI features.

​
Summary

Tracer provides execution-level observability across diverse scientific and compute environments. Its capabilities are defined by what it supports, what it guarantees, and what it deliberately does not require, independent of any specific workflow engine or cloud provider.

Tracer/collect
Learn how execution signals are captured at the kernel level

Tracer/tune
Analyze and optimize pipeline performance and resource usage

Tracer/sweep
Uncover systemwide cloud waste using real execution behavior