Originally published August 26, 2021. This article reflects the state of the market at time of writing. AWS has since announced the deprecation of GameLift managed fleets; the architectural observations below remain relevant to studios evaluating containerized alternatives.

—

For the past few years, AWS GameLift has been a dominant solution for game studios looking for a reliable and highly scalable infrastructure for their game servers. However, containerization has now made it possible for studios to replicate and extend what GameLift was previously touting as their competitive advantage.

AWS GameLift Was There at the Right Time for Game Studios

There's no question that AWS GameLift is one of the most used solutions for big and small game studios. Its dominance can be explained by the real advantage it was offering to studios in a time of need.

If we go back a few years, game studios were being left behind in the great wave of cloudification of infrastructure because of cost issues. Unlike most applications, game traffic can jump anywhere from 2 to 10 times more traffic during a peak hour against the low on an average day. Combining that with the fact that game servers require a good amount of CPU and memory to run consistently, the gaming industry simply could not justify the switch from bare metal servers to using cloud computing because of its inherently higher costs.

AWS GameLift addressed a real need for Technical Directors and CTOs. You could now scale your infrastructure on-demand based on traffic, and pay only for the capacity used at any moment. Simply configure your game server environment that needs to be replicated, request new servers when you see more demand, and close down servers when your player base is going back to school after a sleepless night of playing your awesome game.

For developers familiar with modern infrastructure, those advantages should ring a bell:

• Deployment consistency – CHECK

• Scaling based on traffic – CHECK

• Cost based on actual usage – CHECK

That's essentially what a container does. Why didn't game studios build a container-based infrastructure instead?

Containers simply weren't seeing the adoption rate they receive today when GameLift launched. But things have been steadily changing.

Where Containerization Goes Further Than GameLift

GameLift's managed fleet product runs game server binaries in VMs rather than containers, which limits portability and scaling flexibility. A container-based solution would also benefit from:

• Mobility: take your container and you can run it anywhere, not just AWS.

• Reliability through cattle-like infrastructure: if something's not working, automatically kill it and launch a new one.

• Scaling efficiency: containers can be launched within seconds. GameLift's managed fleet model requires pre-warming capacity, which means maintaining standby servers that aren't actively serving players.

• Easy updates and rollbacks: containers are easily versioned, retaining all configs and dependencies internally. It's easy to have multiple versions running at the same time, and straightforward to roll back in case of issues.

Containers: Capacity scaling follows precisely the player demand curve.

Cloud (GameLift): Extra unused capacity must always be available in case of increasing player demand.

Hybrid Cloud: A cheaper base of bare metal servers is used, sitting idly when demand is low, and extra scaling is done on the cloud, again with extra unused capacity.

Bare metal: Capacity is fixed, and so must be able to handle the peak of players, resulting in a lot of over-capacity for most of the day.

With the Right Tools, Containers Make Things More Streamlined for Developers and Better for Players

Containers are great by themselves, but with the right tools and the right partners you can start seeing some of the forward-looking capabilities that containerization enables:

• Multi-cloud, Multi-edge, or Hybrid: mobility means you can run your containers anywhere. So why not run them everywhere? It's easy to bring your game servers closer to your players with a multi-provider approach, which can result in lower latency and improved user experience. You might even incur cost savings by choosing the most cost-effective location amongst different providers in the same city.

• Global distribution without global costs: Not sure if your game will be successful in the Middle East? Unsure if players from South Korea will find your main character culturally relevant? It doesn't matter. You don't need to set up game servers in the faint hope that some players will join, and incur costs just to test a new market. Containers can be deployed within a few seconds when the players want to play, where they want to play.

• Precise visibility on any deployment: one of the common concerns with cloud infrastructure is its lack of transparency. With containers, studios have full observability on how their application is behaving at any time in terms of resource usage, performance, deployment traceability, and more.

• Automated management: no more regions to manage, no more server provisioning, server self-regulation, which all translates into having fewer people and resources to manage the backend.

So, What's the Future for AWS GameLift?

AWS GameLift addressed a genuine gap in the market, but containerization has since introduced capabilities that weren't available when GameLift launched and that reflect where the industry is moving. AWS Fargate is a container-based solution that could be used for game studios, though cold-start times can vary significantly depending on configuration and image size, and may not meet the latency requirements of real-time multiplayer sessions. Azure and Google have both introduced game server hosting products in the years since GameLift gained traction, though neither has achieved the same level of adoption in the managed game server space. Google Agones takes a containerized approach, though it still relies on fleet-based scaling managed by the studio, which reintroduces some of the operational overhead that container orchestration aims to reduce. Azure's combination of Container Instances and Playfab represents a potential path forward, though the integration between the two for game server orchestration is still maturing.

There are a few solutions emerging that can ease the transition from AWS GameLift to a container-based orchestration solution. The direction of travel for the industry is clear, and studios evaluating their infrastructure options today have more purpose-built containerized alternatives to choose from than ever before.

—

Originally published August 26, 2021. This article reflects the state of the market at time of writing. AWS has since announced the deprecation of GameLift managed fleets; the architectural observations below remain relevant to studios evaluating containerized alternatives.