KOMAL👩‍💻

Interviewer: YOUR SQL QUERY GOT 40X SLOWER AFTER ADDING A FILTER. Before: 150ms. After: 6 seconds. The filter removes 99% of rows. Why?

Interviewer: YOUR SQL QUERY GOT 70X SLOWER AFTER UPDATING STATISTICS. Before: 100ms. After: 7 seconds. No code or index changes. What happened?

One laptop has 8 CPU cores. Another has 16 CPU cores. Yet the 8-core laptop sometimes finishes tasks faster. Does doubling the cores actually mean doubling the performance?

Interviewer: YOUR DATABASE WRITES GOT SLOWER AFTER ADDING 5 INDEXES. Before: 200ms. After: 1.5 seconds. Why can faster reads make writes slower?

Interviewer: YOUR DATABASE INDEX IS USED, BUT PERFORMANCE IS STILL TERRIBLE. Index scan: 5ms. Total query: 4 seconds. Where is the remaining time going?

Why can adding an index make a SQL query go from 200ms to 2 seconds? Shouldn’t indexes always make queries faster?

Why can 100 concurrent users slow down an API more than 10,000 requests sent one after another? Isn’t the total workload what matters?

Why does a 5GHz Wi-Fi network sometimes perform worse than 2.4GHz? Isn’t 5 supposed to be faster?

Interviewer: YOUR DATABASE GOT SLOWER AFTER DOUBLING RAM. Before: 300ms. After: 2.4 seconds. No query changes. What would you investigate?

Interviewer: YOUR DATABASE HANDLES 10K QUERIES/SEC BUT STRUGGLES AT 2K. CPU is low. Memory is stable. Queries haven’t changed. What hidden bottleneck would you investigate?

Your SQL query returns 1 million rows. You add LIMIT 10, but it's still slow. Why? A) Missing index B) Expensive sorting C) Network latency D) All are possible

Interviewer: YOUR SQL QUERY GOT 50X SLOWER AFTER ADDING LIMIT 10. Without LIMIT: 400ms. With LIMIT: 20 seconds. How is that possible?

What makes a man genuinely happy?

Interviewer: YOUR API GOT SLOWER AFTER ADDING MORE SERVERS. 4 instances → 200ms. 12 instances → 900ms. Why?

Interviewer: YOUR SQL RETURNS 20 ROWS BUT READS 2 MILLION FIRST. What would you inspect before rewriting the query?

AI makes coding knowledge easier to access. But architecture still requires trade-offs, not answers. Does AI make senior thinking more valuable?

Interviewer: DELETE removed 5M rows, but the database size barely changed. Why didn’t the storage immediately shrink?

Interviewer: Your AI agent passes every test, but fails only in production. How would you debug a failure you can’t reliably reproduce?