รูปแบบการทำให้แคชไม่ถูกต้อง

Caching helps speed up applications by storing frequently accessed data closer to where it's needed. But what happens when the original data changes?

Cache invalidation is the process of removing or updating cached data when the original source data has changed. It ensures that users always see the most up-to-date information.

Imagine you're viewing a product's price on an e-commerce site. If the price changes in the database but your browser (or an intermediate cache) still shows the old price, that's stale data.

Stale data can lead to incorrect information, bad user experiences, or even financial losses. Effective invalidation is key to preventing this.

The simplest invalidation strategy is Time-To-Live (TTL). Each cached item is given an expiry time. After this time, the item is considered stale and will be removed or refreshed upon the next request.

It's easy to implement but doesn't guarantee immediate freshness if the data changes *before* the TTL expires.

Pros of TTL:

• Easy to implement and manage.
• Automatically handles removal of old data.
• Reduces load on the database periodically.

Cons of TTL:

• Data can be stale for the duration of the TTL.
• Choosing an optimal TTL can be tricky.
• Not suitable for data requiring immediate consistency.

Explicit invalidation means directly removing a cached item when its corresponding source data changes. This ensures immediate freshness.

When an update occurs in the database, the application explicitly tells the cache to delete the affected item(s). The next read request will then fetch the fresh data from the database and re-populate the cache.

This pattern combines the Cache-Aside strategy (where the application manages caching) with explicit invalidation. It's very common.

How it works:

• Read: Check cache first. If not found, fetch from DB, then store in cache.
• Write: Update DB first, then explicitly invalidate (delete) the item from the cache.

In distributed systems, Event-Driven Invalidation uses a Publish/Subscribe (Pub/Sub) model. When data changes, the service responsible publishes an "update" event to an event bus.

Other services or cache instances that hold a copy of that data subscribe to these events and invalidate their local caches accordingly. This decouples services and ensures consistency across many components.

Another approach is to use version numbers or timestamps. Each cached item and its corresponding database record can carry a version.

When fetching data, you can compare the cached item's version with the database's version. If the cached version is older, it's stale and needs to be refreshed. This is useful for optimistic concurrency control as well.

The best invalidation pattern depends on your application's needs:

• Data Freshness: How critical is it for users to see the absolute latest data?
• Update Frequency: How often does the data change?
• System Complexity: How many services share the data?
• Performance Impact: What's the cost of invalidation vs. the cost of stale data?

Often, a combination of patterns is used.

You are designing a system for a real-time stock trading platform. Stock prices update very frequently, and showing stale prices could lead to significant financial issues for users.

Which cache invalidation strategy would be most appropriate to ensure users always see the most up-to-date stock prices?

In this lesson, we explored crucial cache invalidation patterns:

• Time-To-Live (TTL): Simple, time-based expiry.
• Explicit Invalidation: Direct removal upon data change.
• Event-Driven (Pub/Sub): For distributed systems to notify changes.
• Versioning: Comparing data versions/timestamps for freshness.

Choosing the right strategy ensures data consistency and a reliable user experience in your systems.