BUD Delivers Ultra-Fast Performance for Web3 Metaverse Gaming Platform with Amazon MemoryDB for Redis

In BUD, users can get started with built-in geometric modules and interactive props and build detailed 3D interactive experiences from scratch, without the need for any specialized programming or 3D modeling skills. They can also explore millions of experiences ranging from hangout places to battle royale games.

Unlike traditional games, BUD puts greater value on user interactions—comments, likes, and social connections—which in turn generate massive amounts of data. Users can also drop in and out of a gaming session at any time, making for highly concentrated hot data and rapidly changing content such as room attributes in a personalized user environment. Every user action is made possible with a large number of underlying services, and quick response time is expected from all of them, which places exceedingly high demand on the IT infrastructure in terms of performance, stability, and scalability.

It is still early days for BUD as a startup, and they want to keep the underlying infrastructures simple to manage and maintain. The BUD team is focused on delegating management tasks to the backend as much as possible to achieve efficient day-to-day monitoring and maintenance of the system, so that it can focus more on product innovation, service updates, rapid product iteration, and optimization of the user experience.

BUD needed an infrastructure that could support millions of concurrent users for its web3 metaverse gaming platform, deliver ultra-fast performance, high scalability, and was easy to manage. BUD's application requirements include the ability to support highly intensive disk reads and writes, so focusing on when and how to update the memory and cache has always been a top engineering priority for the team.

Storing data in durable transactional logs, MemoryDB offers effective protection against data loss. Whenever an application writes data to MemoryDB, it's committed to the transaction log before the service acknowledges the request. This approach helps achieve scalable read replicas through the multi-AZ transactional log system to achieve data durability, consistency, and recoverability. MemoryDB also enables automatic data replication across multiple AZs, to meet business continuity needs on a wider scale. MemoryDB clusters shard datasets across nodes, with each shard containing one master node and up to five replica nodes. Using read replicas improves system availability, since it provides a backup mechanism for a seamless switchover to the replica nodes should the master node fail. In practice, for better scalability and performance BUD would asynchronously replicate transactional logs and synchronize the read-only replicas with the main cluster to ensure that applications can read data from any node of the cluster.

With MemoryDB at the core, BUD has built a backend that is just right for its user-generated web3 metaverse gaming platform, with support for large-scale and low-latency data storage. MemoryDB delivers microsecond read and single-digit millisecond write response times and can handle millions of transactions per second (TPS) through multi-AZ parallel processing and data pipeline.

BUD is powered by a host of AWS services. Static content such as pictures are stored in Amazon Simple Storage Service (Amazon S3) and distributed through Amazon CloudFront; dynamic content is stored in databases. BUD uses MemoryDB as the primary database to store session data for the game as well as hot data such as maps, likes and reviews, and to manage user profiles. Amazon Gamelift is a dedicated game server hosting solution that deploys, operates, and scales cloud servers for multiplayer games.

BUD breaks down each independent functional module into microservices in an efficient design on the back-end, and then delivers the aggregated data to the user in a summary view on the front-end. Therefore, a single player’s activities may involve many microservices, multiplied by millions of concurrent users making millions of requests per second. The latency of each microservice is magnified and will either make or break the user experience.

MemoryDB is designed to enable BUD’s microservices architecture since it provides ultra-fast performance—with microsecond reads and single-digit millisecond writes— to support its millions of users. MemoryDB also delivers high scalability to support user activities especially at peak hours, with up to 500 nodes and 100 TB of in-memory storage per cluster (250 primary nodes each with one replica).

In addition, Amazon Elastic Kubernetes Service (EKS), a managed Kubernetes service with a secure environment to start, run, and scale Kubernetes in the AWS cloud is helping BUD build and run complex microservice applications with ease.

Online service hosting is another key AWS use case in BUD. In traditional online hosting models, maintenance scripts and tools are often fragmented, numerous, and not readily reusable; version management and Distributed Denial of Service (DDoS) protection are also major challenges. Not to mention the difficulty of balancing cost, efficiency, availability and the many potential points of failure in running a large number of high-concurrency applications.

Amazon GameLift, a dedicated game server hosting solution, helps BUD solve these problems. Its support for auto scaling, multi-AZ, automated disaster recovery and DDoS protection using a logical design simplifies infrastructure management. Other features include visual session management, user management, global deployment, and flexible player matching (FlexMatch).

Steve Zou, Head of Technical Team of BUD, credits Amazon GameLift with liberating the BUD team from the complex and time-consuming tasks of server maintenance, version segregation, session management, and elastic scaling, instead enabling the team to do more. In a web3 metaverse gaming platform like BUD, the notion of “match round” does not apply, as user participation is simply an event in a player-driven universe. Given how BUD differs from massively multiplayer online (MMO) games in player matching rules, AWS helped the BUD team analyze and implement a unique matching system to meet their needs.

If Amazon GameLift were the “support crew” for BUD’s application services, then MemoryDB could be called the “gold standard” for ensuring user experience. The BUD leaderboard records real-time user scores and rankings, which demands frequent data reads and writes. MemoryDB offers the ZSET Redis data type (sorted sets) built-in which allows for flexible processing of up to hundreds of thousands of data requests per second, which makes managing leaderboards a breeze. “MemoryDB enables near-real-time display of rankings with an update frequency as low as 1-2 minutes. This is a big step forward in user experience optimization in 3D metaverse applications,” said Steve Zou.

Using these technologies, the BUD team is now free from having to deal with database performance and server management tasks and can instead fully devote themselves to innovation. Compared with the traditional “relational database + cache” setup, the BUD team has more than quadrupled their R&D efficiency with MemoryDB.

In addition, BUD is also using other AWS services to expand its capabilities, services, and innovation. Amazon SNS and Amazon OpenSearch Service enable secure, real-time search, surveillance and analysis of business and operational data, enabling BUD to create an all-in-one platform covering online service launch, governance and monitoring.

With AWS, BUD will continue to drive innovation in three key areas: lowering system complexity, reducing development time and costs, and optimizing user experience through rapid iteration. “Speed matters the most in creating an innovative web3 metaverse platform, and BUD puts a high premium on system responsiveness.” Steve Zou concluded, “AWS is a critical component of our platform as its wide range of managed services will greatly accelerate our business.”