When your data is stored on many computers (servers) instead of one, we need to make sure that everyone sees the same data — even if they’re asking different servers.
A consistency model is like a rulebook. It says:
“If one person changes something, when and how should others see it?”
Let’s look at the main types.
1. Strong Consistency
What happens:
Everyone always sees the latest update — instantly.
Example:
You change your phone number in a system. Anyone who checks it after that will see the new number, no matter which server they ask.
Best for:
Bank apps, payment systems — where the correct value is very important.
Downside:
It can be slow, because all servers must agree before giving answers.
2. Sequential Consistency
What happens:
Everyone sees things in the same order, but maybe not right away.
Example:
You first change your name, then your address. Others might see both changes later, but they’ll always see the name change before the address change.
Best for:
Systems where the order of actions matters, like event logs.
3. Causal Consistency
What happens:
If one thing depends on another, people always see them in the right order.
Example:
You post a photo, then someone comments on it. Everyone will see the photo before they see the comment.
Best for:
Social media, chat apps — where replies must make sense.
Note:
If two actions don’t depend on each other, they might appear in different orders for different people.
4. Eventual Consistency
What happens:
Everyone will see the right data — just not right away.
Example:
You change your address. It shows up immediately on your phone, but your laptop still shows the old one. After a few seconds, the laptop catches up.
Best for:
Big systems like Amazon or Gmail. Some delay is okay.
Downside:
People may briefly see wrong or old data.
5. Weak Consistency
What happens:
There are no rules. Sometimes you’ll see the update, sometimes not.
Example:
You upload a photo, but sometimes it doesn’t show up — or it disappears and reappears.
Best for:
Temporary storage or low-priority data where correctness isn’t important.
Summary Table
| Consistency Model | What It Feels Like | Best For |
|---|---|---|
| Strong | Everyone sees the latest data immediately | Banking, transactions |
| Sequential | Everyone sees changes in the same order | Logs, order-sensitive systems |
| Causal | Replies always come after original messages | Social apps, chat |
| Eventual | Everyone gets the correct data, eventually | Large websites, online stores |
| Weak | Anything can happen | Caches, non-important data |
Final Words
Think of these models like different levels of “how much can I trust the data I’m seeing?”
- If you need perfect accuracy, choose strong consistency.
- If you need speed and flexibility, eventual or causal consistency is usually good enough.