All posts
Engineering

Self-Host TracePath or Use TracePath Cloud: An Honest Decision Guide

Both options run the same open-source code. The difference is who carries the pager. A concrete decision framework covering cost, ops load, compliance, and scale ceilings.

The short version

Self-hosting TracePath is genuinely self-hosting: the same binary, MIT licensed, no feature gates. TracePath Cloud is the same code operated by the team that builds it. Neither is a upsell of the other, so the question is purely about your constraints.

When self-hosting wins

  • Compliance requires data to stay inside your own VPC or on-prem.
  • You already run infrastructure and have someone on rotation.
  • Your volume is predictable and you want flat costs — one VM plus storage is the whole bill (see our SQLite benchmark for the floor).

When Cloud wins

  • Nobody on your team wants observability infra as a side job.
  • You want ingest, retention, backups, and upgrades to be someone else's SLO.
  • You need HIPAA-mode / BAA territory, where we run dedicated single-tenant stacks.

TODO-content: TODO-content: include the concrete decision table (data residency, ops ownership, cost at 5/50/500 GB per month, retention tiers, enterprise features like SSO and audit logs), link to /cloud for the plan matrix and to the docs billing page for what happens at 80%/100% quota. Add the migration note: you can export and move between them; self-host accepts the same OTLP endpoints. Close with the honest failure modes of each choice (you own backups vs you own a subscription) and the "start self-host, graduate to cloud" path.

Subscribe

Get new engineering posts in your inbox