Database Highload: When to Scale and When to Optimize

on YouTube ↗49m

" summary

Shows how to decide between scaling and optimizing using concrete metrics (CPU%, requests/sec, disk latency/IOPS, cache life) rather than buzzwords. Describes capturing queries with extended events into a table (example: highload) to group by text and find hot spots, then profiling a slow stored procedure (transaction_history) that took 69s vs others at 12s and cutting its cost by ~90% through targeted optimization. Also warns that SAN drivers or network latency can introduce random 1–30ms disk I/O spikes that affect performance.

AI-written summary · about

" chapters (14)

1. 0:00Introduction
2. 1:10Why databases slow down
3. 3:40Scalability markers
4. 7:30When scaling fails
5. 10:44Diminishing returns in scaling
6. 13:18Memory and caching issues
7. 16:08Disk load and latency
8. 21:13Cloud vs bare-metal scaling
9. 24:28Finding load sources
10. 25:53Query tracing demo
11. 30:26Query optimization and load balancing
12. 35:27Scale or optimize decisions
13. 36:52Optimization vs scaling takeaways
14. 38:39Q&A

" related talks

• 47mMy Database Can't Handle the Load. What to Do?Denys Reznyk · Highload fwdays'17
• 49mHighLoad Solutions On MySQLXiaobin Lin · HighLoad++ 2015
• 49mHighLoad Solutions On MySQLXiaobin Lin · HighLoad++ 2015
• 49mHighLoad Solutions On MySQLXiaobin Lin · HighLoad++ 2015
• 58mTroubleshooting CPU Performance Issue for SQL DevelopersPinal Dave · Great Indian Developer Summit 2015
• 53mOptimizing your Azure SQL Database workloadsSanjay Mishra, Denzil Ribeiro, Dimitri Furman · SQLBits 2020
• 54mPerformance Analysis and Optimization of MySQLPetr Zaytsev · HighLoad++ 2013
• 10mBuilding High-Performance Financial Data PipelinesAnandan Dhanaraj · Conf42 Machine Learning 2025