Data is written to cache by the application which is then written to the database. Data is written from cache to database synchronously; meaning no data is returned by the application until the write is completed.
While data is slow to write, it is very fast to read and not stale.
When to Use
- Simplify application logic by offloading data retrieval
- Strong data consistency required between cache and data store
- Any scenario benefiting from a consistent way to access cache
Use Cases
- Financial transactions
- Ledger of transactions must be the same between all parts of the system to prevent misappropriating money
- Inventory management
- Inventory counts must be accurate to prevent over selling to customers
Disadvantages
- New nodes will only cache entities when updated
- Invalidation may be challenging as written data may rarely be accessed/evicted; mitigate by setting a TTL value
- Write-heavy workloads that vastly outpace reads; no reason to store all writes to cache given low read usage
- Poorer write performance