Cache memory (RAM) is fast but limited. Eventually, it gets full. When you need to store new data but the cache is full, you must delete old data. This is Cache Eviction.

The Concept: Time-to-Live (TTL)

Even if the cache isn’t full, data shouldn’t live forever. Time-to-Live (TTL) is a setting that automatically deletes data after a certain period (e.g., 5 minutes, 1 hour).

The Netflix Example

Netflix used to cache subscription plans with a 10-year TTL.

• Problem: When they changed from 4 plans to 5 plans, users still saw the old 4 plans because the cache never expired.
• Fix: They reduced the TTL to hours/days to ensure users see updates.

Eviction Algorithms

When the cache is full, how do we decide who gets kicked out?

1. Least Recently Used (LRU)

“If you haven’t used it in a while, you probably won’t use it soon.”

• How it works: Removes the item that was accessed longest ago.
• Use Case: Most common strategy. Good for social media feeds (recent posts are hot, old posts are cold).

  %%{init: {'theme': 'dark'}}%%
graph LR
    subgraph Cache ["LRU Cache (Size 3)"]
        A["A (Oldest)"] --> B["B"]
        B --> C["C (Newest)"]
    end
    New["New Item D"] -.->|Evicts A| A
    style A fill:#ff9999,stroke:#333
    style New fill:#99ff99,stroke:#333

2. Least Frequently Used (LFU)

“If you rarely use it, get rid of it.”

• How it works: Tracks how often an item is accessed. Removes the one with the lowest count.
• Use Case: Netflix (Bollywood movies might be less watched than Hollywood blockbusters globally, so Bollywood gets evicted first in US servers).

3. First In First Out (FIFO)

“Oldest data leaves first.”

• How it works: A simple queue. First item added is the first one deleted.
• Use Case: Time-series data logs.

4. Most Recently Used (MRU)

“Delete the newest item.”

• Use Case: Rare. Example: “Tinder” (you don’t want to see the same person you just swiped on again).

5. Random Replacement

“Just delete something.”

• Use Case: When you don’t care or want to avoid overhead.

Strategy Comparison

Conclusion

There is no “perfect” strategy.

• For most web apps, LRU is the safe default.
• For content platforms (Netflix), LFU might be better to keep popular content hot.