Home · Tech in 60 Seconds

Transcript

1. Can one database handle a billion users?
2. Not easily. So you split the data across many databases. That's sharding, like library branches.
3. One: the shard key. Like an author's last name, every book has one.
4. Two: how to split. Range groups A to H. Hash spreads them evenly.
5. Three: finding it. A router reads the key and sends you to the right branch.
6. What if one branch gets slammed? Just build another?
7. That's a hotspot. Go from three branches to four with simple hashing, and 75% of the books move.
8. So we use consistent hashing. Put branches on a ring, and only about a quarter moves.
9. Sharding splits the data. Replication copies it. In practice, you use both.
10. Splitting is easy. Re-splitting is hard.
11. Recap: shard key, range or hash, router.
12. Changing a shard key means moving all your data. So pick it from your most common queries.

Sources

1. Microsoft Learn, Azure Architecture Center, "Sharding pattern" (2026-04-02)
2. MongoDB Manual, "Sharding" (확인 2026-10-27)
3. MongoDB Manual, "Shard Keys" (확인 2026-10-27)
4. MongoDB Manual, "Ranged Sharding" (확인 2026-10-27)
5. MongoDB Manual, "Hashed Sharding" (확인 2026-10-27)
6. Vitess Docs, "VTGate" (확인 2026-10-27)
7. Apache Cassandra Documentation, "Dynamo" (architecture, 확인 2026-10-27)
8. MongoDB Manual, "Replication" (확인 2026-10-27)