5.2.0-beta1
Pre-releaseWhat's new
This release supports server-assisted, client-side caching, and is currently beta grade.
Client-side caching is available within UnifiedJedis, JedisPooled, JedisCluster, etc classes via implementation of ClientSideCache class, with only RESP3 protocol. It is recommended to use a ClientSideCache implementation with TTL (time-to-live). We have included two implementations based on Google Guava and Caffeine libraries.
How to try Client-Side Caching
- Install Jedis 5.2.0-beta1
- Choose and install a caching library: Google Guava or Caffeine
- Use the following code example to get started:
import redis.clients.jedis.*;
import redis.clients.jedis.csc.*;
class CacheExample {
public static void main() {
HostAndPort node = HostAndPort.from("localhost:6379");
JedisClientConfig clientConfig = DefaultJedisClientConfig.builder()
.resp3() // RESP3 protocol
//.user("myuser") // Redis server username (optional)
//.password("mypass") // Redis user's password (optional)
.build();
ClientSideCache clientSideCache;
// Uncomment one of the following lines to use the corresponding cache backend
// GuavaClientSideCache clientSideCache = GuavaClientSideCache.builder().maximumSize(10_000).ttl(100).build();
// CaffeineClientSideCache clientSideCache = CaffeineClientSideCache.builder().maximumSize(10_000).ttl(100).build();
UnifiedJedis client = new UnifiedJedis(node, clientConfig, clientSideCache);
// JedisPooled client = new JedisPooled(node, clientConfig, clientSideCache);
// JedisCluster client = new JedisCluster(Collections.singleton(node), clientConfig, clientSideCache);
client.set("foo", "bar");
client.get("foo");
client.get("foo"); // cache hit
client.del("foo");
client.close();
}
}
It is highly recommended to use a ClientSideCache implementation with TTL.
Both of our provided GuavaClientSideCache and CaffeineClientSideCache have TTL support and use a default value when not set. It is discouraged to use ttl(0)
in these.
It is also a good idea to keep the idle connections busy to get more and updated notifications. It can be done easily using pool config:
GenericObjectPoolConfig<Connection> poolConfig = new ConnectionPoolConfig();
poolConfig.setTestWhileIdle(true); // ConnectionPoolConfig by default does this.
// It is still shown here for better understanding.
This pool config can be used as follows:
JedisPooled client = new JedisPooled(node, clientConfig, clientSideCache, poolConfig);
JedisCluster client = new JedisCluster(Collections.singleton(node), clientConfig, clientSideCache, poolConfig);
It is possible to limit or ignore commands and/or keys for client side caching. For example, if we want to ignore some keys based on their prefix, we can define a ClientSideCacheable:
final String IGNORE_PREFIX = "PREFIX_TO_IGNORE";
ClientSideCacheable isCacheable = new ClientSideCacheable() {
@Override
public boolean isCacheable(ProtocolCommand command, Object... keys) {
for (String key : (String[]) keys) { // assuming we'll only execute methods with String keys
if (key.startsWith(IGNORE_PREFIX)) {
return false;
}
}
return true;
}
};
This ClientSideCacheable can be a parameter for ClientSideCache. In our provided implementations, it can be:
GuavaClientSideCache clientSideCache = GuavaClientSideCache.builder().cacheable(isCacheable).build();
CaffeineClientSideCache clientSideCache = CaffeineClientSideCache.builder().cacheable(isCacheable).build();
It is also possible to create client-side caching enabled client object using URL/URI with proper query params. Supported params are:
cache_lib
(caching library) - required; can be eitherguava
orcaffeine
cache_max_size
(maximum no of commands) - optionalcache_ttl
(time-to-live, in seconds) - optional
For example:
JedisPooled client = new JedisPooled("redis://myuser:mypass@localhost:6379/?cache_lib=guava&cache_max_size=10000&cache_ttl=100");