Loading data from the database every time user makes a similar request can increase the API response time. In such cases Cache can be used.
What is a Cache?
Cache is a temporary data storage layer which is much faster than database. Cache works like map, which has a key and value every time you enquire for that key it’s corresponding value is returned. The benefits of having separate Cache will create better performance and would also reduce the DB hits or Database calls.
User makes a request to the application, now application is going to check for the given request/key data is present inside Cache or not, if Cache contains the data then read from the Cache. If Cache doesn’t contain data then read from Database and save this data into the Cache. This process is called read-through cache.
When to use a Cache?
Cache should be used when the data is read frequently but modified infrequently. Data inside Cache is stored inside volatile memory and thus it is not created for persisting data. If Cache server restarts then the data present inside memory will be lost. Therefore, you can not just rely on Cache for storing all your data and you should always store data inside persistent data stores.
Every cache that you configure should have an expiration policy. Once cached data expires it will be removed from the cache. If expiration policy is not in place then data inside Cache will be stored permanently unless Cache memory becomes full. Your expiration date should should not be too short as this will cause Cache misses and database calls will increase significantly. At the same time it should not be too long as the cached data can become stale.
It is important to keep your database and cache in sync. Inconsistency can happen because of database being modified and not updating cache in the same transaction.
You should have cache servers across different data centers as if a single cache server exists then there can be “A Single Point of Failure” (SPOF) and it will stop entire system from working. One more thing that can be done to present failure would be to configure memory more than that is required.
What will happen if the Cache memory becomes full?
Once Cache memory is full, any new items which are added to it will cause removal of existing items. This process is known as Cache Eviction. Least-recently-used (LRU) is one of the most popular Cache eviction policy.