Busiverse
273 posts
Busiverse
@Busiverse1
Turn ideas into operating businesses. A distributed universe powered by blockchain and AI. System Design Foundations book: https://t.co/4t14I0zqFw
Katılım Mayıs 2025
5 Takip Edilen1 Takipçiler
Interview answer: “I’ll start reasonably normalized, then denormalize for performance-critical queries—while defining how I keep duplicates consistent.” book.busiversehq.com
English
Rule of thumb: start normalized (core truth tables), then denormalize selectively for hot read paths (cached aggregates, immutable snapshots). book.busiversehq.com
English
Trade-off:
Normalized → cleaner writes, more joins.
Denormalized → faster reads, more complex writes + sync risk.
book.busiversehq.com
English
Classic denormalization moves:
• store product_name/thumbnail on order_items
• store comment_count (avoid recounting on every page load)
book.busiversehq.com
English
Normal forms, interview-level:
• 1NF: atomic values
• 2NF: non-key depends on full key
• 3NF: non-key depends only on key
book.busiversehq.com
English
Denormalization = duplicate on purpose (store derived/duplicate fields). Result: faster reads + simpler queries (often fewer joins). book.busiversehq.com
English
Normalization = store each fact in ONE place (reduce duplication). Result: fewer inconsistency bugs + better data integrity. book.busiversehq.com
English
Normalization vs Denormalization: one is “clean truth”, the other is “fast reads”. Knowing when to use each is a system design superpower. book.busiversehq.com
English
