Caching

Pasted image 20230902110723.png

There are a few ways to update our cache:

Cache eviction

When no one used the cache and there is no eviction happens, cache can be stale.
To deal with it we can set a TTL and then try to expire the item
We can:
- Passively expire an item — when an item got accessed it checks for TTL to expire
- Actively expire item — have a thread to go through and query the expiration cache

Pasted image 20230906130004.png

Cache Aside:
- Application will handle the cache miss, when there is a cache miss, application will read the db and update the data itself
- Pros:
  - Fine-gain control of how to update the cache
- Cons:
  - Need to handle cache eviction and expiration
Read Through:
- The cache itself will handle the cache miss, if there is a cache miss, the cache will fetch the db and update itself ==without going through application==.
- Pros:
  - Simplicity
- Cons:
  - Less control over the implementation

Write through: write into both database in the same time
- Pros:
  - high resilient
  - less cache miss
  - no data stale
- Cons:
  - double the storage
Write around: write into database first and then the cache can update later (either through cache miss or something)
- Pros:
  - no data stale
  - less memory storage
- Cons:
  - Higher chance of cache miss
Write back: write to the cache first and let the cache asynchronously update our database
- Pros:
  - less cache miss
- Cons:
  - Single point of failure (cache)
  - DB might have stale data