Two levels of cache (Redis + Caffeine)

Question

When profiling an application it came up that Redis is impacting the execution times because there are many sleeps in threads. I need to implement two levels of cache or think about solution of this problem.

I would like to have two levels of caches:

L1 - local for each instance of deployment,
L2 - cache global for all instances of same deployment,

The solution that I came up with is:

Create two CacheManagers (CaffeineCacheManager and RedisCacheManager),
Initialize same caches for each cache manager,
Use annotation @Caching with cacheable={} to use two caches,


    @Caching(cacheable = {
            @Cacheable(cacheNames = CacheConfiguration.HELLO_WORLD),
            @Cacheable(cacheNames = CacheConfiguration.HELLO_WORLD, cacheManager = "cacheManagerRedis")
    })
    public String generate(String name)
    {
        log.info("  Cached method call...");
        return helloWorldService.generate(name);
    }

The structure of classes is similar to: CachedService (annotations here) -> NonCachedService

The problem I am facing:

I wanted to have it working in flow (yes - works/n - not working):

[ y ] data is fetched and then cached to both caches Redis and local - this works
[ y ] if data exists in local cache do not move it to redis - this works
[ y ] if any of caches contains the data it will be fetched from cache
[ n ] if data exists in Redis, move it to local - this does not work

Modification of @Caching annotation to have put={} where it would put values into local cache is making whole cache not working.


    @Caching(cacheable = {
            @Cacheable(cacheNames = CacheConfiguration.HELLO_WORLD),
            @Cacheable(cacheNames = CacheConfiguration.HELLO_WORLD, cacheManager = "cacheManagerRedis")
    }, put = {
            @CachePut(cacheNames = CacheConfiguration.HELLO_WORLD),
    })
    public String generate(String name)
    {
        log.info("  Cached method call...");
        return helloWorldService.generate(name);
    }

Do you know any spring-ready solutions to work with two levels of cache?
I've read about local caching with Redis but it does not mean anything similar to my case (it's just the standard redis use case),
I am left only with double-layered services structure to achieve this goal? Similar to CachedLocal -> CachedRedis -> NonCached

user_3574368 · Answer

For anyone else looking for this, I was able to implement the CacheInterceptor suggested by Ankit.

Example:

public class RedisMultiCacheInterceptor extends CacheInterceptor {

@Autowired
private CacheManager caffeineCacheManager;

@Override
protected Cache.ValueWrapper doGet(Cache cache, Object key) {
    //Get item from cache
    var superGetResult = super.doGet(cache, key);

    if (superGetResult == null) {
        return superGetResult;
    }

    //If retrieved from Redis, check if it's missing from caffeine on local and add it
    if (cache.getClass() == RedisCache.class) {
        var caffeineCache = caffeineCacheManager.getCache(cache.getName());

        if (caffeineCache != null) {
            caffeineCache.putIfAbsent(key, superGetResult.get());
        }
    }

    return superGetResult;
}

}

Two levels of cache (Redis + Caffeine)

1 Answers