Point-in-Time Recovery in the Gaming Industry

Scenarios

Databases may experience faults such as data corruption, loss, or accidental deletion. To ensure services are running properly, you need to restore a database to a normal state before the faults occurred. Traditional databases adopt a periodic backup policy. That is, data is restored when the system is faulty. Data restoration takes a long time. As a result, customer services are severely affected.

Solution Overview

Point-in-Time Recovery (PITR) allows you to restore the database to a particular point in time if data is lost or corrupted due to misoperations or accidental deletion.

Some game players may exploit vulnerabilities to duplicate equipment and currency, leading to unfairness. Traditional databases are backed up once a day, making it difficult to restore data to a specific point in time. With PITR of GeminiDB Redis API, you can specify a specific time point for data restoration. Data can be restored within 5 minutes at least.

Advantages

PITR of GeminiDB Redis API maintains your data from past timestamps and does not affect data snapshots. If there is a fault, data can be restored to a specified point within 5 minutes. Therefore, GeminiDB Redis AP is widely used in industries such as gaming and finance.

• Backup tasks are not affected, and services are running stably.

  PITR does not affect data backup and access.

  GeminiDB Redis API enables you to create snapshots by recording the file system status instead of by copying files. File metadata (such as data block information and addressing information) at the current moment is stored to generate snapshots. Therefore, services are not affected during snapshot creation.

• Data is restored in minutes regardless of its volume.

  PITR snapshots can be stored on your local PC and do not need to be uploaded to cold storage media. Therefore, data replication and migration are not involved, and data can be restored anytime.

  Even hundreds of GB of data can be restored within 5 minutes. After being restored to a specified point in time, data can also be restored multiple times to a state before or after that point in time.

• GeminiDB Redis API has better backup performance than open-source Redis

  Open-source Redis uses copy-on-write (CoW) to effectively enable snapshot persistence in its multi-thread architecture. When Redis calls fork() to create a child process, its parent process will be blocked for hundreds of milliseconds. As a result, jitter will occur. CoW may cause memory overuse. If a large number of writes are performed while the parent process is forked, memory is severely wasted or even OOM occurs (memory usage < 50%). PITR frees you from copying or migrating data, so services are not affected. Snapshots can be quickly created, and data can be restored stably and securely.

Solution

For details about how to enable PITR of the GeminiDB Redis API and restore data to a specified time point, see Restoring to the Original Instance Using PITR.