GAMEPERF01-BP01 Evaluate game server resource requirements and scalability needs
Evaluate server requirements against your scalability needs to verify that you are selecting a hosting option that both meets your requirements and provides optimal performance.
Level of risk exposed if this best practice is not established: High
Implementation guidance
When selecting the appropriate hosting option for your game servers, consider the following factors:
Game server resource requirements
Assess the CPU, memory, network, and storage requirements of your game server processes to determine what your game consumes. Do not overlook networking; each frame requires CPU cycles to receive player actions, update the state of the game, and send it back to the player. Offloading packet processing can free up CPU for core game functions. Networking is the foundation for smooth and responsive game play so testing it early in the process defines a baseline performance profile for a game.
A first person shooter game might have high actions per second
which the CPU needs to quickly move off to the network which may
favor compute optimized C family instances, while a turn-based
strategy game which can spend more CPU cycles processing per turn
may need increased memory from R family instances to locally store
and update the state of the game on the server before sending it
back to players. Use a data-driven approach like
the
Utilization Saturation and Errors (USE) Method
Scalability and elasticity
Evaluate how quickly and smoothly each hosting option can scale to meet player demand without compromising performance. Consider the level of automation and flexibility required for your game's workload to maintain a smooth gaming experience during peak times. A game server might scale quickly by increasing utilization through adding additional game server processes on the same instance, where a game backend may scale slower based on the rising active user count and games being played. Your fleet should scale with demand to minimize cost while facilitating minimal wait time for the players to get into game. Review Amazon EC2 Spot Instance Advisor to gain insight into cost effective available capacity for game server fleets.
Implementation steps
-
Evaluate game server resource requirements for CPU, memory, network, and storage to select suitable instance types, considering game-specific performance needs such as high network throughput for FPS games or memory optimization for turn-based strategy games.
-
Compare different hosting options such as containers, instances, bare-metal, and managed services by analyzing performance data using frameworks like the USE method. Use these insights to make better decisions about your system architecture.
-
Design fleets for scalability and elasticity, leveraging tools like EC2 Spot Instance Advisor to optimize costs while facilitating quick scaling to meet player demand during peak times.
