Tair (Redis® OSS-Compatible):Use Bloom filters to manage game event push notifications

How it works

A Bloom filter is a probabilistic data structure that answers membership queries with minimal memory. It returns one of two results:

• Definitely not in the set — the player has not received the notification; send it.

• Possibly in the set — the player may have received the notification; skip it.

The trade-off: false positives are possible, meaning a small number of players might miss a notification. For most game event scenarios, this is an acceptable trade-off for the memory and performance gains.

Why Bloom filters for this use case:

Prerequisites

Before you begin, ensure that you have:

• A Tair (Enterprise Edition) instance with the endpoint and port number

• The instance password

• Java 8 or later

• Maven

Manage notifications with a Bloom filter

The following example uses three operations to control notification delivery:

import redis.clients.jedis.*;
import redis.clients.jedis.UnifiedJedis;

public class TairBloomFilterDemo {
    // Replace with your instance endpoint, port number, and password.
    static HostAndPort hostAndPort = new HostAndPort("r-bp1y****svonly41srpd.redis.rds.aliyuncs.com", 6379);
    static JedisClientConfig config = DefaultJedisClientConfig.builder().password("tw:Da***3").build();
    static UnifiedJedis unifiedJedis = new UnifiedJedis(hostAndPort, config);

    private static final String BLOOM_KEY = "activity_popup";

    /**
     * Creates the Bloom filter with a 1% false positive rate and capacity for 50,000 player IDs.
     */
    public static void createBloom() {
        try {
            unifiedJedis.bfReserve(BLOOM_KEY, 0.01, 50000);
        } catch (Exception e) {
            e.printStackTrace(); // Handle connection timeouts and other errors.
        }
    }

    /**
     * Returns true if the player has not yet received the notification.
     */
    public static boolean shouldShowPopup(String playerId) {
        try {
            return !unifiedJedis.bfExists(BLOOM_KEY, playerId);
        } catch (Exception e) {
            e.printStackTrace(); // Handle connection timeouts and other errors.
            return true;
        }
    }

    /**
     * Records that the player has received the notification.
     */
    public static void updatePopupState(String playerId) {
        try {
            unifiedJedis.bfAdd(BLOOM_KEY, playerId);
        } catch (Exception e) {
            e.printStackTrace(); // Handle connection timeouts and other errors.
        }
    }

    /**
     * Sends a notification to the player if they haven't received one yet.
     */
    public static void handlePopup(String playerId) {
        if (shouldShowPopup(playerId)) {
            System.out.println("Push a notification to the player: " + playerId);
            updatePopupState(playerId);
        } else {
            System.out.println("Player " + playerId + " has already been notified.");
        }
    }

    public static void main(String[] args) {
        createBloom();
        String playerId = "player123";

        handlePopup(playerId); // Sends the notification on first login.
        handlePopup(playerId); // Skips the notification on subsequent logins.
    }
}

Expected output:

Push a notification to the player: player123
Player player123 has already been notified.

Choose a false positive rate

The error_rate parameter in bfReserve controls the trade-off between accuracy and memory. A lower rate improves accuracy but increases memory consumption.

In the example, the Bloom filter is created with a 1% false positive rate and a capacity for 50,000 player IDs. Adjust these values based on your expected player base and accuracy requirements.

False positives and notification delivery

A false positive means the filter incorrectly reports that a player has already received a notification when they have not. The affected player misses one notification. For most game event scenarios, this is acceptable. If missing a notification is not acceptable—for example, for time-sensitive rewards—consider supplementing the Bloom filter with a secondary confirmation step.

Tair (Enterprise Edition) compatibility

The Bloom filter provided by Tair (Enterprise Edition) is compatible with the Redis Bloom filter and uses the same commands. See TairBloom commands for the full command reference.