Tracer Documentation
Ready to get started?
Step 1: Install Tracer
Ready to get started?
Step 1: Install Tracer
Explore Tracer by features
Pipeline Intelligence
Pipeline-level intelligence built from kernel signalsSee the real behavior of every stage, tool, and binary, exactly the way scientific pipelines run.
Performance intelligence for every runReveal bottlenecks and opportunities for right-sizing with a complete view of CPU, GPU, memory, disk, and I/O utilization at every step.
Catch idle compute and hidden waste instantlyTracer detects idle instances, ghost GPUs, stalled jobs, and silent retries the moment they happen.
Cost & Resource Governance
True cost attribution (AWS/GCP 1:1)See cost by pipeline, tool, user, team, or cost center. No tagging or manual accounting required. Just accurate spend mapped to scientific work.
Instance rightsizing recommendationsOptimize your cloud spend with data-driven recommendations based on actual resource usage patterns.
Debugging & Root Cause Analysis
Root-cause insight for failures and slowdownsNo logs needed to pinpoint the exact tool, syscall, or resource bottleneck behind OOMs, I/O stalls, network slowdowns, and silent pipeline failures.
Generate logs for tools that don’t logTracer produces structured timelines and events even for tools that output nothing, which makes it feasible for you to debug and reproduce.
Automatic tool & syscall detectionDiscover every binary, script, container, and syscall involved in a workflow automatically.
Platform Compatibility & Security
Framework-agnostic runtime visibilityNextflow, Snakemake, WDL, CWL, Python, Ray, Slurm, Batch, if it runs on Linux, Tracer sees it, and your pipelines stay exactly as they are.
Runs secure and lightweight at the kernel levelTracer collects system telemetry, never scientific data, code, or secrets. It adds near-zero overhead, pipelines stay fast and private.
Dive deep into all our features in the features section.
Bring your own HPC environment
Tracer supports a wide array of environments to fit your infrastructure and workflow needs.Your environment is not shown here? Check out all our environments.
Observability should be framework agnostic
Tracer works with any framework and instantly gives you detailed signals your framework can’t surface. Pipeline optimization is no longer a chore, it naturally emerges when you can see what happens inside your workflows.Is your framework is not shown here? Check out how to integrate Tracer with your framework.
Tutorials
Learn how to use Tracer as efficiently as possible. We have created a set of tutorials where we guide you through the most common use cases to help you get started with Tracer.You can also dive deeper into how the technology on which Tracer was built:
How It Works
High-level overview.
High-level overview.
eBPF
Revolutionary Linux kernel tech.
Revolutionary Linux kernel tech.
Tracer Agent
Architecture and capabilities
Architecture and capabilities
How Tracer fits into your stack
Learn more about how Tracer can extend the capability of your current stack and give you a deeper view into your pipelines.Seqera
Observability for Nextflow.
Prefect
Just beyond orchestration.
Dagster
Scheduled runs visibility.
Grafana
Customizable observability.
To see how Tracer complements other tools go to Comparisons.
Key Use Cases
Deep Observability for Scientific Pipelines
Monitor pipelines end-to-end.
Team-Wide Observability and Collaboration
Beyond individual pipelines.
Cost and Resource Optimization
Reduce cloud compute costs.
Debugging and Failure Analysis
Accelerated through data.
About Tracer
Tracer was founded in 2023 by Vincent Hus and Laura Bogaert after experiencing firsthand how difficult it was to set up, run, and debug scientific pipelines at scale.They two set out to build what would become the world’s first observability platform purpose-built for scientific computing.