Design Pastebin

Question 1

Who is pasting and who is reading?

Accepted Answer

Every design starts with the caller. Put it on the board even when it feels obvious — the read path and the write path both begin there.

Question 2

What does the browser actually talk to?

Accepted Answer

Something has to receive the paste, generate the id, and decide what a link is allowed to return. That is a service, and it is where every rule in this system lives.

Question 3

Where does the text live?

Accepted Answer

A paste has to survive a restart, so it needs a durable store. At 10KB a paste this fits comfortably in a database — big enough to matter, small enough not to need object storage.

Question 4

A paste goes viral and is opened ten thousand times. What happens?

Accepted Answer

Pastes never change once written, and a small number of them get most of the traffic. That is the easiest caching case there is — read once, serve from memory after, and never worry about invalidation.

Question 5

Your client reads the store directly. What stops it reading every paste?

Accepted Answer

A link is meant to grant access to one paste. A direct connection to the store grants access to all of them, and there is nowhere to check expiry either.

Question 6

How many pastes a second is 1M a day?

Accepted Answer

1,000,000 / 86,400 seconds is about 12 a second. Doing this sum early is what tells you whether you are designing for a laptop or a data centre — here, a laptop.

Question 7

And how many reads a second?

Accepted Answer

At 10:1 that is about 120 reads a second. Small — but the ratio is the point, because it tells you which path to spend your effort on.

Blueprint — system design