Explain like I'm five · Database

What is a Distributed Database?

A distributed database, told as a story about splitting one giant warehouse into regional branches that keep each other in sync. The deeper distributed-systems theory lives in the SRE field.

every order nationwide piles onto one single warehousea far-away warehouse means slow delivery too

Every order nationwide piles onto one single warehouse.A far-away warehouse means slow delivery too.

just copy everything to every branch, right?but what if the copies' stock counts stop matching?if two branches edit at once, the copies' counts drift apart

Just copy everything to every branch, right?If two branches edit at once, the copies' counts drift apart.

Seoul branchBusan branchDaejeon branchDistributed DB = a warehouse network split across branches that keep each other in syncthe nearest branch answers requests fast, yet from outside it looks like one warehouse

A distributed database is a warehouse network split across branches.The nearest branch answers requests fast, yet from outside it looks like one warehouse.

Split by regionthat's sharding / partitioning
Copy for safetyif one branch goes down, another answers
Agree with a voteseveral branches vote on one value
Accept a little delaythe copies may catch up a little late
CentralizedDistributedfast joins, slow far-away requestsone building down = everything downone rulebook makes consistency easyfast answers from the nearest branchone branch down, the rest keep goingcross-branch joins and consistency get hardweigh this trade-off — speed vs. consistency and join cost — when choosing

Choosing centralized or distributed trades response speed against consistency.The cost of cross-branch joins belongs in that choice too.

sharding, replication, and consensus are the basics of distributed DB modelingdistributed-systems theory itself goes deeper in the SRE field

Sharding, replication, and consensus are the basics of distributed DB modeling.Distributed-systems theory itself goes deeper in the SRE field.

In one breath

A distributed database stores one logical database split across many nodes, using techniques like replication and consensus to make it look like a single database.

From a DB-modeling angle, this is about choosing a sharding key, picking sync vs. async replication, and the rising cost of cross-node joins and transactions. The deeper distributed-systems theory — CAP theorem, consensus algorithms — is covered in more depth over in the SRE field.

When grown-ups say it

Sharding
Splitting goods by branch. Choosing the split key is covered further in this same field. → splitting an overgrown warehouse into zones
Replication
Keeping a duplicate box at each branch. Covers sync vs. async replication. → the world of duplicate boxes
Consensus
Several branches voting on one answer. Covers algorithms like Raft/Paxos. → deciding by a vote among many
CAP Theorem
The law that three properties can't all be perfect at once. The consistency/availability/partition trade-off. → the law you can't have all three of at once
Eventual Consistency
Copies eventually match, even if a bit late. Copy counts can differ for a moment. → the story of eventually matching up
Data Locality
Handling a request at the nearest branch. The closer the branch, the faster the answer.
Distributed Transaction (2PC)
A transaction slip spanning several branches. The slip only finalizes once every branch agrees.
NoSQL
A warehouse without fixed shelf slots. Loosens the structure to make distribution easier.
Next story
Finding by Several Tags at Once →
Want another term as a picture book? Request →