How Plaid Migrated 234 Databases to TiDB with a Six-Person Team

Migrating hundreds of databases sounds like a job for a large, specialized team. Plaid did it with six engineers, most without deep database expertise, moving 234 databases across 100 services from Amazon Aurora MySQL to self-hosted TiDB in two and a half years.

REGISTER HERE: https://www.pingcap.com/event/how-plaid-migrated-234-databases-to-tidb/

In this webinar, Plaid Platform Architecture Lead Zander Hill shares the leadership playbook behind the project: How his team turned resource constraints into an advantage, earned organizational buy-in for a multi-year re-platforming effort, and built compounding automation that made each migration faster and more reliable than the last.

You’ll discover how Plaid categorized services by business criticality to avoid over-engineering, chose controlled cutovers with 60 to 120 seconds of writer downtime over zero-downtime complexity, and frontloaded their riskiest Tier 0 services to de-risk the entire program. The result: Zero downtime across three TiDB upgrades in 2025, each costing only four engineer weeks compared to 12 on Amazon Aurora.

REGISTER HERE: https://www.pingcap.com/event/how-plaid-migrated-234-databases-to-tidb/

What you’ll learn:

• How to scope a large-scale migration for a small team by aligning to business tolerances, not engineering perfection.
• Why hiring for ownership and operational excellence matters more than deep database pedigree.
• How incremental automation, including runbooks, SDKs, and async job runners, compounds into a decisive speed advantage.
• A risk-frontloading strategy that surfaces critical flaws early and builds organizational confidence.
• How to sustain cross-functional goodwill and low attrition across a 2.5-year infrastructure project.

REGISTER HERE: https://www.pingcap.com/event/how-plaid-migrated-234-databases-to-tidb/