GAMEPERF04-BP01 Monitor game server processes to detect issues

You might run multiple game server processes per instance to efficiently utilize the resources on your game server instances. If so, design your architecture so that an individual game server process hosting a game session cannot cause adverse impact to other game sessions hosted on the same instance. Use metrics to understand how game placement and game mode type can impact the performance of game server instances. Incorporate a mix of low load (lobby, shop, or single player tutorial) and high load (ranked, multi-player, or high skill gameplay) processes to avoid hot spotting the game server instance.

Level of risk exposed if this best practice is not established: High

Implementation guidance

Monitor the player experience through client side and server-side metrics by collecting telemetry for ping time and jitter, frame drops, API response time, errors and successful game loop completion. Correlate time stamps for these events with player support issues and server logs to identify performance bottlenecks. Tools like Dtrace, ftrace, uperf, and eBPF can be used for deep investigation and analysis of system performance.

Implement monitoring of the limited resources available to your game server instances so that you can generate alerts when individual game server processes are breaching pre-determined resource budget thresholds. When thresholds are breached, you may want to configure your game server software to dump relevant system and game server logs out to durable storage, such as a central logging solution, so that your game server engineers can investigate this behavior. Additionally, your game server instance should be configured to report metrics from each of the game server processes running on the instance so that you can monitor these individual game server processes in addition to the overall metrics for the game server instance.

For example, GameLift provides metrics for monitoring game sessions, which can be augmented with custom game-specific metrics and logs collected using the Amazon CloudWatch Agent which you can configure on your game server instance. Your metrics can be viewed in CloudWatch or exported to other tools such as Amazon Managed Grafana which is integrated with Single Sign-On to make it straightforward to access metrics by users who may not have access to the Management Console. Refer to the following best practices for managing logs and metrics using Amazon GameLift, which also provides support for viewing individual game session logs.

Implementation steps