Delicious: Hardware, Database, and the 1 GB Rule

At this point, Joshua has Delicious online and it is starting to go viral. Millions of bookmarks are arriving, and he is hitting his first real walls: Hardware and Storage.

2003 Hardware: A Different Reality

To appreciate the scale, I had to look back at what "power" actually meant in the early 2000s. Even in 2007, a mid-range PC came with:

• 128 MB of RAM: That is megabytes, not gigabytes.
• 40 GB Hard Disk: I remember shopkeepers warning that this was "too much" and you'd never fill it in your life.
• Core 2 Duo CPU: Having two cores was a massive luxury.

Joshua's 2003 laptop was likely even weaker than that.

Rule #1: Dedicated Servers Only

Joshua's first scaling bottleneck wasn't actually his code - it was his lifestyle. Because he was serving the site from his personal laptop:

1. Nightly Shutdowns: When Joshua went to sleep and closed his laptop, Delicious went down for the whole world.
2. Resource Contention: If he tried to play a game, the website slowed to a crawl for everyone else.

The lesson here is simple: For a system to scale, it must run on Dedicated Servers - machines that stay on 24/7 and do nothing but serve requests.

The MVP Database Schema

To support the basic features (Login, Add Bookmark, View List), we only need two simple tables:

1. Users: id, username, password.
2. User_Bookmarks: user_id, url.

ExpandDelicious MVP Database Schema

The Engineering of Data Types

One thing I realized is that your choice of data types sets your ultimate ceiling.

Why use an 8-byte Integer (BigInt) for user_id? A standard 4-byte integer can store up to 4.2 billion values. That sounds like a lot, but there are over 5 billion internet users. If you use a 4-byte integer, your system will literally break when the 4,294,967,296th user tries to sign up.

The Gangnam Style Bug: In 2014, the Gangnam Style video became the first to cross 2.1 billion views. YouTube was using a 4-byte signed integer to store views. When it crossed the limit, the number "overflowed" and became negative. They had to rewrite their entire infrastructure to use 8-byte integers.

The Storage Math: The 1 GB per Day Rule

Joshua is now receiving 1 million new bookmarks every single day. Here is the math that started the countdown:

1. Row Size:
   • user_id (8 bytes)
   • url (averaging ~1,000 bytes for long URLs)
   • Total: ~1 Kilobyte (KB) per bookmark.
2. Daily Growth:
   • 1,000,000 bookmarks x 1 KB = 1 Gigabyte (GB) per day.

At this rate, Joshua's 40 GB hard drive - the one that was supposed to "last a lifetime" - will be 100% full in just 40 days.

He is on a clock. In the next part, I will show you the two choices he had: making his machine stronger or adding more of them.

The Essentials

1. Production systems must run on Dedicated Servers that are on 24/7 - running a live service from a personal laptop is the first bottleneck most new developers overlook.
2. Data type choices are architectural decisions - using a 4-byte integer for a user ID that will eventually exceed 4.2 billion is a system-breaking mistake, as YouTube's Gangnam Style overflow proved.
3. Simple storage math reveals real deadlines: 1 million bookmarks per day at ~1 KB each = 1 GB/day, meaning a 40 GB disk fills up in 40 days - this is how engineers think about capacity planning.

Further Reading and Watching

• Hussein Nasser - Database Indexing, Storage, and I/O - A practical deep dive into how databases store rows on disk and why data type choices have physical consequences.
• PostgreSQL Documentation - Data Types - The official reference for integer sizes, storage requirements, and the exact limits that caused bugs like the Gangnam Style overflow.