GVERSE, a Sandbox-mode Metaverse Gaming Platform, Migrates 1,450 Tables to the Cloud in 26 Days

With the lowering of development barriers, sandbox games, with their interactivity, may become the key to ushering in the "3.0 era content creation". Unlike traditional RPG (role-playing game) and SLG (simulation game), the development tools built into sandbox games are easy to learn, so even a "newbie" with no programming or modeling foundation can easily become a creator. Players become the main source of content output, and everyone can become a developer.

GVERSE’s Blockman Go platform allows players to create several different types of games for other players to play, and also provides rich social interaction features, making it a UGC (User Produced Content) platform. UGC platforms such as "Roblox" and "Minecraft" currently on the market have considerable volume and superior market prospects. At the same time, there are some unique problems with such platforms.

Zheng Hengguo, technical partner and technical director, GVERSE, points out that compared with standard online games, UGC platform games are characterized by the comprehensive coverage of players and large and trivial data volume. Therefore, it puts the configuration of servers under great stress. In addition, since games were often distributed overseas and serves players in many countries and regions simultaneously, GVERSE's services had to be scattered and deployed on partners' local IDCs and other cloud service providers. It not only made the system extremely complicated, but also brought additional management and operational costs.

After its hosting contract ended with a third-party vendor, GVERSE decided to move its data back and manage it. At the same time, it also wanted to solve business fragmentation, a pain point, completely. Zheng mentions that GVERSE had extensive contact with Amazon Web Services before. Having used Amazon Web Services ’s dynamic CDN (content delivery network), network acceleration and big data services, he was impressed by the extensive layout and expansive stability of Amazon Web Services 's global multi-nodes during the process. Because of that, GVERSE continued working with Amazon Web Services for its database migration.

There were only 26 days to migrating its data off disparate locations. Every day of the project was a "peak moment.” Amazon Web Services brought in a project management team to assist with this. It also developed a targeted migration plan for GVERSE. The migration covered 1,450 large, medium, and small tables, over 2TB of MySQL database, and 20TB of data with over 600 million objects.

"First, this migration is to collate game data originally scattered across many different IDCs and clouds without omission to ensure no loss. Second, it is to ensure data verification in the migration process. If there are data errors, we need to recover them in time to preserve data. Finally, this migration involves complex product categories, covering Amazon S3, Amazon RDS, Amazon DynamoDB and so on," Zheng says. "This is heavy and complicated work. ‘Heavy’ meaning large volume of data. ‘Complicated’ is reflected in need to consider many aspects such as detail combing, compliance, version and activity configuration."
 

Since Amazon DynamoDB has auto-scaling capabilities to reliably scale to millions of concurrent user requests while ensuring consistent low latency of no more than 10 milliseconds, GVERSE primarily used the Amazon DynamoDB and Amazon Lambda to complete the migration.

The table data volume of GVERSE was vast and numerous. It involved 1,450 tables of different sizes, which is quite rare. Among these, the user forms contained the total amount of user data, and import and export time was estimated to be 7 hours. If GVERSE used the traditional "Double-Write" method, it might take several months to migrate. In addition, due to the UGC property of the customer's product, each opened mini-game is stored in a separate table, and there are more than 1000 mini-games (created by gamers) on the whole platform. However, as the activity level varies, so the challenge of a large number of tables and many small tables arose. To make matters worse, GVERSE required the database migration to be completed within a few hours on the cutover day to protect the user experience. Therefore, it would make the migration task even more challenging.

GVERSE and Amazon Web Services replicated most extensive tables using CDC solution (Change Data Capture), and the remaining tables were migrated in batches and categories. They migrated smaller tables first, and then different types of tables were migrated in batches by using the Amazon CLI Command line of Amazon DynamoDB S3 export and Amazon DynamoDB S3 import. They made it just in time before the cutover date.

GVERSE's MySQL database is about 2TB. It chose Amazon Aurora to host its data for ease of management. Amazon Aurora's fully managed features significantly saves operation and maintenance costs. Its features of automatic execution of standardized database clustering and replication solved the biggest problems in database configuration and management.

For smaller data buckets with a large number of small files under 40K, the previously used CRR (Cross-Region Replication) method was no longer efficient. Instead, Amazon Web Services migrated this data using the Amazon DTH solution. It has external features such as multi-source data support, serverless architecture (no upfront costs, pay-per-use), and intuitive ease of use. It is especially suitable for transferring many files of different types and sizes from multiple IDCs or clouds to native Amazon Web Services. The implementation of DTH has enabled the easy migration of over 600 million shredded files. The duration was shortened from one month in the CRR method to three days, perfectly matching GVERSE’s needs.

To allow players to create and generate games in a risk-free UGC way, after the migration was completed, GVERSE also deployed Amazon Web Services security sevices such as Amazon Guarduty RDS Protection, Amazon Inspector, Amazon Security Hub, and Amazon WAF to protect user data from infrastructure threats. This helped mitigate potential risks and gave its customers confidence that all its businesses were running on Amazon Web Services.