31.2 How big is this, really?

Step 2: Estimates (5 minutes)

Using Chapter 2's moves (times, divide, times) on the agreed numbers. Assume about 500 bytes per stored link and 5 years of links.

What	Working	Result
New links a second	100 million ÷ 2.6 million seconds in a month	≈ 40 a second
Redirects a second	40 × 100	≈ 4,000 a second average
Peak redirects	4,000 × 3 (assumed peak factor)	≈ 12,000 a second
Links in 5 years	100 million × 12 × 5	6 billion
Storage	6 billion × 500 bytes	≈ 3 TB
Code length	base62: 62⁶ ≈ 56.8 billion > 6 billion	6 characters is enough; 7 gives room to spare (3.5 trillion)
Cache for hot links	redirects a day ≈ 333 million × 500 bytes × 20% hot	≈ 33 GB, fits in one Redis node's memory

What the numbers say, out loud:

"40 writes a second is small. One database handles that. Writes aren't the hard part."
"12,000 reads a second at peak is where the effort goes: caching."
"3 TB fits comfortably on one database server, with replicas for safety. No sharding needed on day one."

Interviewer pushback:

"What if a link goes viral and gets a million clicks in a minute?" → That's a hot key (Chapter 13). The cache handles most of it; for extreme cases I'd add a small in-memory cache on each app server, or let a CDN cache the redirect.
"Why 7 characters if 6 is enough?" → Headroom. If our estimate is off or the product grows, 7 characters gives about 60 times more codes for one extra character.

Memory code: "Writes are tiny, reads are the job." The numbers tell you where to spend the next 30 minutes.