Understanding Bloom Filters and Their Implementation with Google Guava and Redis

Bloom filters are space‑efficient probabilistic data structures that can test whether an element is a member of a set, offering fast queries with a small false‑positive rate.

A Bloom filter consists of a bit array and K independent hash functions; adding an element hashes it K times and sets the corresponding bits to 1.

Because bits may be set by multiple elements, a query may return true for an element that was never added, which is the inherent false‑positive characteristic; however, a negative result is always correct.

Google's Guava library provides an in‑memory Bloom filter implementation. The article shows how to add the Guava dependency and create a BloomFilter<Integer> in a Spring service, populate it from a database, and expose a REST endpoint to check membership.

<dependency>
    <groupId>com.google.guava</groupId>
    <artifactId>guava</artifactId>
    <version>21.0</version>
</dependency>

@Service
public class BloomFilterService {
    @Autowired
    private UserMapper userMapper;

    private BloomFilter<Integer> bf;

    /**
     * Create Bloom filter (default 3% error rate)
     */
    @PostConstruct
    public void initBloomFilter() {
        List<User> userList = userMapper.selectAllUser();
        if (CollectionUtils.isEmpty(userList)) {
            return;
        }
        // create Bloom filter
        bf = BloomFilter.create(Funnels.integerFunnel(), userList.size());
        for (User user : userList) {
            bf.put(user.getId());
        }
    }

    /**
     * Check if an id might exist in the Bloom filter
     */
    public boolean userIdExists(int id) {
        return bf.mightContain(id);
    }
}

@RestController
public class BloomFilterController {
    @Autowired
    private BloomFilterService bloomFilterService;

    @RequestMapping("/bloom/idExists")
    public boolean ifExists(int id) {
        return bloomFilterService.userIdExists(id);
    }
}

Redis offers a distributed Bloom filter via the ReBloom module. After launching a Docker container with

docker run -d -p 6379:6379 --name bloomfilter redislabs/rebloom

, commands such as bf.add, bf.exists, bf.madd, and bf.mexists can be used. The filter can be pre‑allocated with bf.reserve to control error_rate and initial_size.

# Docker installation
[root@localhost ~]# docker run -d -p 6379:6379 --name bloomfilter redislabs/rebloom

# Basic usage
[root@localhost ~]# docker exec -it bloomfilter /bin/bash
root@container:/data# redis-cli -p 6379
127.0.0.1:6379> bf.add urls www.taobao.com
(integer) 1
127.0.0.1:6379> bf.exists urls www.taobao.com
(integer) 1
127.0.0.1:6379> bf.madd urls www.baidu.com www.tianmao.com
1) (integer) 1
2) (integer) 1
127.0.0.1:6379> bf.mexists urls www.baidu.com www.tianmao.com
1) (integer) 1
2) (integer) 1

# Reserve a filter with custom parameters
127.0.0.1:6379> bf.reserve user 0.01 100
OK

The article compares the two implementations: Guava's filter lives in JVM memory and is lost on restart, making it unsuitable for distributed scenarios, while Redis's filter is persistent and scalable but incurs network latency.

Finally, a practical use case is presented: filtering non‑member users in a membership lottery, where the Bloom filter eliminates 99 % of non‑members before a database lookup confirms actual membership.