Preventing Cache Penetration in Spring Boot Using Redis and Bloom Filters

Cache penetration is a problem where frequent requests for non-existent keys bypass the Redis cache and directly hit the relational database. To prevent this, a Bloom Filter guard layer can be implemented in front of Redis and PostgreSQL. The Bloom Filter is a compact probabilistic data structure that can efficiently check if an element is definitely not in a set or might be in a set.

In the code, a `CachePenetrationGuard` class is defined with a `BloomFilter` instance. The filter is configured to handle 500,000 expected insertions with a 1% false positive rate. The `registerKey` method adds an account ID to the filter, while `mightContain` checks if an account ID is likely to be present.

In the `AccountService`, before querying Redis or PostgreSQL, the Bloom Filter is consulted. If the account ID is not likely to be in the filter, the request is immediately rejected, saving the database from unnecessary lookups. If the filter indicates the account ID might be present, Redis is checked. If the data is found in Redis, it is returned. If not, the data is fetched from the database, cached in Redis, and then returned.

Combining Bloom Filters with Redis' TTL jitter can effectively shield backend databases from cache penetration and traffic spikes in production.