How Amazon GameLift Achieved 100 Million Concurrent Users: A Scalability Case Study

Jump to

Amazon GameLift, a managed game server hosting solution by AWS, has set a new benchmark in online gaming scalability by supporting up to 100 million concurrent users (CCU) for a single game. This groundbreaking achievement demonstrates the platform’s ability to dynamically scale backend resources, ensuring seamless gameplay experiences even under extreme traffic conditions. This case study explores how Amazon GameLift reached this milestone, its core features, and its impact on the gaming industry.

The Challenge of Scaling for Modern Games

Game developers face immense pressure during game launches as millions of players flood servers simultaneously. The backend infrastructure must scale instantly to handle unpredictable spikes in traffic without compromising player interactions. Failure to do so can result in server crashes, long wait times, and poor user experiences.

Chris Byskal, GM of AWS Game Services, highlighted the magnitude of this achievement: “While most popular games today top out around 14 million CCU, Amazon GameLift’s ability to support 100 million CCU demonstrates scalability several times beyond current industry standards.”

Amazon GameLift’s Scalability Features

Dynamic Autoscaling

Amazon GameLift dynamically scales backend resources based on player demand. During benchmark tests, the platform added up to 100,000 players per second and spun up over 9,000 new compute instances per minute. This ensures that games can accommodate massive traffic surges without delays or disruptions.

Global Infrastructure

With capacity available across 23 AWS Regions and 9 Local Zones, Amazon GameLift offers low-latency gameplay by placing servers closer to players. This global reach ensures smooth experiences for users worldwide.

Efficient Resource Allocation

GameLift optimizes resource usage by autoscaling fleets of virtual machines (VMs) to match demand. For example:

  • To support 10 million CCU with an average of 12 game sessions per VM, approximately 83,333 VMs are required.
  • Autoscaling minimizes idle capacity, reducing operational costs while maintaining readiness for player surges.

Preparing for Scale: Key Considerations

Estimating Capacity Needs

Developers can calculate VM requirements using projected CCU and game session parameters. For instance:Total VMs=CCU(players per session)×(sessions per VM)Total VMs=(players per session)×(sessions per VM)CCU

For a game supporting 10 million CCU with 10 players per session and 12 sessions per VM:Total VMs=10,000,00010×12=83,333Total VMs=10×1210,000,000=83,333

Game Session Allocation

Efficiently allocating players to game sessions requires balancing several factors:

  • Latency: Assigning players to servers with the lowest latency.
  • Cost: Prioritizing lower-cost capacity options.
  • Capacity Health: Ensuring servers are responsive and capable of handling new sessions.
  • Avoiding Overload: Preventing excessive game sessions on a single VM to maintain performance.

Scaling Demonstrations

Battle Royale Scenario

In a simulation for a battle royale game with 100 players per session:

  • The system scaled to support 100 million CCU.
  • Despite the higher player count (111,200 new players per second), Amazon GameLift maintained efficient session allocation.
  • Over 90% of players were placed into games within three seconds.

High-Throughput Scenario

In another test simulating a head-to-head fighting game with three-minute sessions:

  • The system handled over 2,700 game session allocations per second.
  • With lower compute requirements (50 sessions per VM), only 10,000 VMs were needed for one million players.

These scenarios highlight Amazon GameLift’s ability to adapt to diverse gaming requirements while maintaining performance and cost efficiency.

How Amazon GameLift Works

Integration and Setup

Developers integrate their game servers with Amazon GameLift by uploading the server executable as a Build and creating Fleets in AWS Regions. Fleets combine hardware resources with uploaded builds to run game sessions efficiently.

Autoscaling Policies

Autoscaling policies ensure optimal resource utilization by targeting specific thresholds for available game session capacity. For example:

  • Initial policies target 40% available capacity during launch surges.
  • Post-launch adjustments reduce this threshold to minimize costs while maintaining availability.

Game Session Queues

Game Session Queues route players to the best available fleet based on latency and cost considerations. This ensures that players are matched with the most suitable servers for their location and gameplay needs.

Testing Scalability

To simulate real-world conditions during game launches:

  1. Load Test Generator: Simulates player traffic by creating game session requests at increasing rates.
  2. Player Simulator: Connects simulated players to allocated game sessions while measuring latency and performance.

During tests for a hypothetical launch:

  • The system scaled from zero to over 91,000 VMs in minutes.
  • Player requests were processed without delays or disruptions, even as traffic exceeded initial estimates by tenfold.

Lessons from Extreme Scalability

Amazon GameLift’s benchmark tests underscore several key takeaways:

  1. CCU is Just the Beginning: While CCU is an important metric, factors like VM deployment speed and session allocation throughput are equally critical for scaling success.
  2. Global Coverage Matters: Deploying fleets across multiple Regions minimizes latency but requires careful planning to avoid fragmenting the player base.
  3. Optimized Autoscaling: Dynamic scaling ensures cost efficiency while maintaining readiness for sudden traffic spikes.

Conclusion

Amazon GameLift’s ability to support up to 100 million concurrent users sets a new standard for scalability in online gaming. By leveraging dynamic autoscaling, global infrastructure, and efficient resource allocation strategies, it empowers developers to deliver seamless experiences even under extreme conditions.

For developers preparing for their next big launch, Amazon GameLift offers a proven solution capable of handling unprecedented levels of player traffic while optimizing costs and performance. Contact an AWS representative today to learn how Amazon GameLift can elevate your gaming infrastructure.

Read more such articles from our Newsletter here.

Leave a Comment

Your email address will not be published. Required fields are marked *

You may also like

Categories
Scroll to Top