Skip to content
BENIBIX Learn

Strong Consistency vs Eventual Consistency

People think strong and eventual consistency differ only in speed. With strong consistency, every read sees the newest write at once. It does not matter which copy you ask. With eventual consistency, a copy may show old data for a short time, then catch up. Strong is slower, and it may turn calls away when the network splits. Eventual stays fast and up, but you may see an old value.

Development🏗️ Architecture

Strong Consistency

Each read shows the last write. It does not matter which copy answers.

Strong consistency means that once a write is done, every read after it sees the new value. It does not matter which copy of the data a reader asks. To do this, the copies must agree before the write is confirmed. That takes extra messages, so writes are slower. If the network splits, such a system may turn requests away rather than give old data.

Example
Emma moves 50 euros between her bank accounts. Her phone and her laptop both show the new balances the moment the transfer is confirmed.
Why it matters
Money, stock counts and bookings cannot show old numbers. Strong consistency stops two people from buying the last seat.
Full entry
Development🏗️ Architecture

Eventual Consistency

Copies of data may differ for a moment. Then they all match.

Eventual consistency means copies on different machines may differ for a short time. In the end they all match. A change is saved on one machine first and then copied to the others. While it travels, a reader may see the old value. Once no new changes arrive, every copy catches up. DNS works this way: a changed address can take hours to reach everyone.

Example
Emma changes her profile photo. For a few seconds, a friend in another country still sees the old one, then the new photo appears.
Why it matters
It keeps systems fast and online around the world. The price is a short spell of old data.
Full entry

In one line

Strong Consistency
Each read shows the last write. It does not matter which copy answers.
Eventual Consistency
Copies of data may differ for a moment. Then they all match.