Neon to ClickHouse: Postgres Migration Success

Unlocking Postgres Performance

The article presents a compelling case for migrating from Neon to ClickHouse Managed Postgres, highlighting tangible benefits in reliability, performance, and cost. The use of real-world examples from diverse companies like Shipsidekick, socialpruf., and Infinitas Learning lends significant credibility. The detailed explanation of the migration process using ClickPipes, including the step-by-step guide and the mention of handling schema changes and connection pooling with PgBouncer, is highly valuable for potential adopters. The immediate performance improvements, particularly the reduction in p99 latency from seconds to milliseconds, are impressive and directly address common pain points of scaling relational databases. The elimination of network transfer fees for teams already using ClickHouse for analytics is a significant economic advantage.

However, the article could benefit from a more nuanced discussion on potential limitations or edge cases. While Neon's serverless architecture is presented as a source of issues, the article doesn't deeply explore the specific architectural choices within Neon that might have contributed to these problems. It would be beneficial to understand if these issues are inherent to Neon's serverless model or specific to certain configurations or usage patterns. Additionally, while the migration is described as straightforward, the mention of Neon's deprecation of Azure regions and the requirement to move to entirely new servers for major Postgres version upgrades suggests underlying platform stability concerns that might extend beyond just performance under load. A more in-depth comparison of the operational overhead and feature sets of both managed Postgres offerings, beyond just the migration experience, would also strengthen the analysis.

Key Points

• Teams are migrating production Postgres workloads from Neon to ClickHouse Managed Postgres due to issues with reliability under load, inconsistent performance growth, rising infrastructure costs, and network transfer fees.
• Post-migration benefits reported include improved reliability, faster and more consistent query performance (p99 latency dropping from seconds to milliseconds), fewer operational issues, and lower costs.
• The migration process using ClickPipes, a feature of ClickHouse Managed Postgres, is highlighted as fast and straightforward, handling schema migration, initial load, and change data capture (CDC) for synchronization.
• ClickHouse Managed Postgres's architecture, utilizing local NVMe storage, is credited for delivering faster and more predictable query performance by eliminating network-attached storage bottlenecks.
• For existing ClickHouse users, co-locating Postgres on the same platform eliminates network transfer fees when streaming data changes.

---

📖 Source: From Neon Postgres to ClickHouse Managed Postgres