Building on AWS | AWS RDS Latency: How to Handle High IOPS Spikes
00:00 - The Scenario: Shopify's Flash Sale Database Performance
01:45 - Understanding Read vs. Write IOPS
02:50 - Deep Dive: GP3 SSD Storage Explained
04:00 - Analyzing Production Logs: The Latency Cliff
06:25 - Architecture Overview: AWS RDS MySQL on GP3
08:40 - What is P95 Latency? (Why it matters for SLOs)
10:00 - SSD vs HDD: Why Random I/O matters for Databases
11:45 - The OS Waiting Queue: When Hardware Hits the Limit
13:30 - Log Analysis: Correlating IOPS Spikes with Latency
16:00 - The "Throughput Cliff": Why 5% more traffic causes 200% latency
18:00 - Diagnostic Matrix: Compute Bound vs. I/O Bound
20:30 - The Incident Response Framework: Triage, Hypothesis, Mitigation
22:00 - Hypothesis Validation: Storage Queuing confirmed
23:50 - The Fix: Provisioning IOPS on RDS (Online Modification)
25:30 - Conclusion & Key Takeaways