Definitions

The AWS Well-Architected Framework is based on six pillars: operational excellence, security, reliability, performance efficiency, cost optimization, and sustainability. AWS provides multiple core components that allow you to design state-of-the-art architectures for your game workloads. In this section, we will present an overview of key definitions.

For the purposes of this paper, a game architecture encompasses the backend technical infrastructure required to build and operate a game. Some games may not have social, multiplayer, or other online features and may not require the use of certain aspects of backend technical infrastructure that are described in this paper. For a detailed discussion of the different types of workloads that are frequently deployed to support a game architecture, see Scenarios.

The AWS Cloud infrastructure is built around Regions and Availability Zones.

Depending on the characteristics of your game, you may want to deploy certain components of your game architecture into multiple Regions for reasons such as improving performance for players, or to deliver customized experiences for players depending on their location.

There are many different types of games, and the backend technical infrastructure required to support a game differs depending on the type of game being developed. For example, popular types of games may include first person shooters (FPS), role playing games (RPG), massively multiplayer online games (MMOG), battle royales (BR), sports games, puzzle games, and more. There are also different game interaction modes that influence the architecture of the game, such as turn-based and simultaneous play, with different performance characteristics.

Games are developed to be played on one or more gaming systems including desktop, web, mobile, consoles, and newer interaction modes such as augmented reality (AR), virtual reality (VR), and game streaming solutions. Games commonly support cross-system gameplay, which means that players can save their game progression and resume gameplay on other systems, as well as initiate gameplay sessions with players on other systems.

Video game monetization enables game publishers to generate revenue using different strategies like advertising, digital and retail-based game purchases, in-game purchases of downloadable content (DLC) known as microtransactions, and through required paid subscriptions to play the game. Some of the most common key performance indicators (KPIs) in the games industry include:

Gaming system

Video games are developed to be played on a gaming system that provides client input controls, graphics, client software (known as the game client) and hardware, and in some cases system-exclusive features to support gameplay.

Gaming systems are generally separated into the following categories:

In addition to the previously mentioned systems, there are also nascent systems that are still relatively new and growing and have much smaller market share compared to the more predominant systems. Examples of gaming systems in this category include AR, VR, and game streaming, sometimes referred to as cloud gaming.

Game streaming involves rendering the gameplay in the cloud and streaming to thin client, typically a browser. Game streaming allows a player to play a game that is entirely hosted remotely, typically in the cloud by a game streaming service provider. In game streaming, the player connects to a cloud-based game through a browser or a thin client provided by the cloud gaming service provider (gaming system).

Game server

Game servers represent one of the most important aspects of the compute infrastructure for your game. Game servers, sometimes referred to as dedicated game servers, are used when developing a multiplayer game or when server authoritative processing of gameplay events is required. The game server is at the center of the game architecture, serving as the location where the core logic runs, which includes managing player and game state as well as managing the interactions between the connected game clients and the game server. The game server is usually one of the most performance-sensitive aspects of a game architecture because it is responsible for processing the inputs from a player's game client and properly distributing it to other connected players in real-time. A poorly performing game server impacts the overall performance of the game experience. Therefore, you should optimize game server performance and provide sufficient capacity, especially at game launch or peak gameplay periods.

For the purposes of this document, a game server or game server instance refers to the compute, such as a virtual machine, that hosts one or more game server processes. A game server process represents a single instance of your game server build hosting a game session, which is an instance of your running game that players can connect to through a player session. For this reason, we often refer to game server process or game session interchangeably due to the implied one to one relationship between a game session and the game server process that is hosting it. In AWS, there are multiple options for compute to host game servers, which provide access to scalable cloud-based capacity through elastic provisioning of resources.

Amazon EC2 provides cloud-based virtual servers, known as instances, with support for multiple versions of Linux and Windows. You can create instances and manage them directly like another server or virtual machine. Typically, multiple game server processes are deployed to an instance to improve efficiency and reduce costs. Amazon EC2 is a good choice for game servers if you desire the most control over the compute infrastructure.

Amazon GameLift provides a fully managed solution for dedicated game server hosting in the cloud as well as additional features such as matchmaking with GameLift FlexMatch. GameLift provides a layer of abstraction on top of Amazon EC2 to make game server management straightforward and is available in most AWS Regions so that you can host game servers close to players to reduce latency, achieve high availability, and significantly reduce costs by using Spot Instances. While GameLift can be integrated into existing game backends, it is especially useful for game developers who do not want to develop their own game server management and matchmaking solutions and prefer a solution that is managed by AWS and can scale as their game grows.

Amazon Elastic Container Service (Amazon ECS) is a fully managed container orchestration service that you can use to run Docker-based containers. You can also use Amazon Elastic Kubernetes Service (Amazon EKS) to run Docker-based containers that are built using Kubernetes. Using container technologies such as those provided by Amazon ECS and Amazon EKS can assist you to improve your compute utilization by efficiently packing many game server processes or other game application instances into an EC2 instance.

The use of containers can also improve developer productivity by hosting applications using the same Docker image operating runtime used by developers on their local machines during development. You can further reduce operational overhead by using AWS Fargate, which is a serverless compute solution for running containers and is compatible with both Amazon EKS and Amazon ECS. Fargate is best suited for use cases where you want to run game servers in containers without responsibility for operating the underlying instances that the containers run on.

You can use AWS Outposts to run AWS infrastructure and services in a data center or on-premises facility, which can enable games to run in on-premises environments and AWS using the same infrastructure to support a hybrid cloud adoption strategy. AWS Local Zones serve as extensions of AWS Regions that allow your game servers and other latency-sensitive workloads to run more closely to your players or development teams. Additionally, to reduce global network latency for your game servers, you can use AWS Global Accelerator to improve performance for player traffic to your game servers.

AWS Lambda is a serverless compute service that runs code without provisioning or managing servers, which makes it useful for asynchronous game server use cases such as turn-based games or those that have lightweight compute requirements, a small codebase, and where gameplay functionality can be designed using a stateless microservices architecture. It is important to keep in mind that Lambda functions run on an event-driven, per-request basis, rather than running a long-running game server process. Lambda provides the most runtime abstraction of the options in this paper because the underlying application is readily available for developers to choose from to host their code.

When selecting your approach for game server hosting, consider various requirements including operational overhead, legacy codebases, performance requirements, and scale. EC2 instances and containers are good options for legacy codebases, as they require the smallest change to move to the cloud, and you can use EC2 instances to dedicate the resources of a compute instance, while containers can make management and high utilization simpler to achieve. Serverless functions offer the highest level of abstraction, which you can use to define code that only runs in response to events, which can reduce costs.

Game client

The game client represents the software and hardware device that the player uses for playing a game. The game client provides the software for translating the player's inputs into messages that are sent to a server for processing, and it is responsible for handling the incoming responses from the server and rendering outputs such as graphics to the player. In real-time networked multiplayer games, the game client usually maintains a persistent network connection to a game server for the duration of a gameplay session to reduce network latency and minimize processing time. However, the game client may also interact using REST with a game server or backend services.

Messaging

There are typically three primary categories of messages in games: 

A common strategy for sending and receiving these types of messages is to use publisher-subscriber and asynchronous processing architecture patterns. AWS provides several services that can assist you to implement messaging in your game.

Live game operations (Live Ops)

Live game operations (Live Ops) is a style of game management and operations that treats a game as a live service and continually delivers new features, updates, promotions, in-game events, and improvements to the launched game to improve the experience for the player community.

Traditionally, games were delivered as products rather than services, and new content and features were frequently incorporated into subsequent releases or sequels rather than into the launched product. With a Live Ops approach to game management, a game operations team can launch a game and maintain an engaged player community through experimentation, promotions, in-game events, and innovation to keep players entertained.

Although this approach has the benefit of unlocking new player engagement strategies and delivering recurring revenue streams, it requires more operational expertise. For example, to implement a successful Live Ops strategy, a developer may need to integrate with cloud services or operate their own backend technical infrastructure. They also need an effective way to identify and respond to issues that arise in the game, or within the player community, that can negatively impact the player experience.