Due Diligence Report: Database Topology & Scalability Risk
A structured assessment of whether a platform's database topology survives peak load, not just its documentation.
Executive Summary
This assessment evaluated the primary data layer of a multi-tenant platform against synthetic peak load, rather than relying on architecture documentation or team interviews alone. The platform's stated capacity assumptions were tested directly against production-equivalent traffic patterns.
Findings
The current shard key distribution showed uneven load concentration under synthetic peak conditions, with a subset of shards approaching saturation well before the platform's stated capacity ceiling.
Recommendation: Re-evaluate shard key selection before the next major customer cohort onboarding; stage a topology migration with rollback capability rather than a forced cutover.
Automated failover was present but had not been tested against a cascading, multi-node failure scenario, only single-node failure.
Recommendation: Run a cascading-failure game day before relying on the failover path for a production incident.
Connection pool limits were configured for average load, not burst load, creating a plausible cascade trigger during traffic spikes.
Recommendation: Resize pool limits against burst percentiles, not average traffic, and add explicit backpressure handling.
Conclusion
The platform is not currently at risk under normal operating conditions, but the margin between normal load and failure threshold is narrower than its documentation implies. The recommended topology work is estimated at low operational risk if staged, and material risk if deferred past the next scaling milestone.
Frequently Asked
What does a database topology risk assessment actually test?
It tests shard distribution, failover behavior, and connection handling against synthetic peak and burst load, not just whether the documented architecture looks correct on paper.