IllFonic Lowers Latency and Reduces Server Costs by 70% on Amazon GameLift

IllFonic is an independent video game developer with offices in Golden, Colorado; Tacoma, Washington; and Austin, Texas. It creates unique, immersive online multiplayer games. And to provide a seamless gaming experience, IllFonic must quickly match similarly skilled opponents and keep latency low for players around the world. But the peer-to-peer (P2P) network model the company was using to support its multiplayer games created inconsistent experiences and often led to game interruptions and long player wait times.

Already using other services from Amazon Web Services (AWS), IllFonic migrated one of its past titles to Amazon GameLift, a fully managed dedicated server hosting solution that deploys, operates, and scales cloud servers for multiplayer games. This enables IllFonic to deliver the lowest latency possible, reduce player wait times, and significantly cut costs. “To run a successful online multiplayer game today, you need fast servers on demand that can automatically scale when you need them to,” says Charles Brungardt, CEO and cofounder of IllFonic. “On Amazon GameLift, we lowered latency for our players. So we got that—and more.” So successful was the experience that in 2020, IllFonic developed and launched its latest title entirely on AWS.

Founded in 2007, IllFonic has over 100 game development and technology experts. The nimble developer has collaborated on well-known games to create unique asymmetrical multiplayer experiences. To match players in a past title, IllFonic used a P2P network model—an information technology infrastructure in which two or more devices directly connect to share resources without a central server, with one device acting as the server host.

But the P2P model presented problems for the team and its players. Hosts that suddenly dropped from a game, whether by intention or poor internet connection, caused the game to disconnect for all remaining players in the match. Additionally, players can more easily cheat with a P2P network, and hosts can have latency advantages over the other players. Before using the P2P model for player matchmaking, IllFonic had used self-managed dedicated servers for a title’s initial digital release on Steam, the largest digital distribution platform for computer games. Dedicated servers would minimize game interruptions and cheating and would level the playing field. But IllFonic’s self-managed servers had created slow service and had consumed too much time and money to operate.

So the company chose Amazon GameLift to manage its dedicated servers. “Since we were already familiar with the services offered by AWS, having adopted them piece by piece for previous releases, we looked at Amazon GameLift to offload the burden of self-managing server instances on Steam, shift away from the P2P model on console, and move the game entirely to fully managed dedicated game servers in the cloud,” says Paul Jackson, chief technology officer at IllFonic.

In October 2017, IllFonic migrated its Steam player base to dedicated servers deployed and managed by Amazon GameLift in only 2 months. To the satisfaction of players, IllFonic reduced average matchmaking wait times from several minutes to under 10 seconds. Players also experienced lower ping—that is, lower lag in service—while playing. And now if the host drops out of the game, the remaining players are unaffected and can continue the session. The successful move convinced IllFonic to also migrate the console player base to Amazon GameLift.

Later, IllFonic developed its newest title entirely on AWS and launched the game in April 2020. The ability to offload its infrastructure management to AWS meant that IllFonic’s small team of three engineers could spend almost all its time working on game-related tasks rather than server maintenance. “Launch is a lot of pressure,” says Brungardt. “But on AWS we were able to sleep well at night, knowing that the project was working at scale and things were holding up.” To optimize matchmaking in the game, IllFonic implemented Amazon GameLift FlexMatch, which matches players based on skill, latency, and custom criteria. However, the IllFonic team had experienced technical issues while first implementing it: the system had placed players in presorted buckets without running rule set comparisons first, leading to incompatible matches. IllFonic was able to remedy this issue with the help of the AWS support team. “We reached out to AWS and had a call that night,” Jackson says. “The team told me about a feature that would fix the issue but that wasn’t technically released yet, and that alleviated a lot of the problems—all those failures went to zero within a couple weeks.”

By integrating Amazon GameLift FleetIQ and Amazon Elastic Compute Cloud (Amazon EC2) Spot Instances, IllFonic cut the costs of running its servers by 70 percent. The automatic scaling feature of Amazon GameLift enables IllFonic to automatically add or remove compute capacity to meet fluctuating workloads. And IllFonic uses the Amazon GameLift FleetIQ algorithm to make real-time predictions for which Spot Instances will be best suited to host new game sessions. This capability not only saves on costs but also directs players onto instances less likely to have game session interruptions. The combination of Amazon GameLift FleetIQ and Spot Instances saves the company money and time while maintaining high availability for players. IllFonic can now reinvest these resources into other areas of the company, and its team can focus on game development and improvement. “We were skeptical of statistics that showed that instance recalls would occur less frequently than on-premises hardware failure,” explains Brungardt. “We thought it was too good to be true, but since launch in April 2020, we’ve only seen one session ever be interrupted and have seen zero support tickets reporting servers being removed and causing session disconnection.”

Also improving IllFonic’s ability to provide a better player experience is the suite of AWS services it uses to track and manage player data and profiles—information IllFonic uses to manage player progression and inform the development of its games. “Data is incredibly important to our running the game,” explains Jackson. “For example, we use it for player reporting, so if players report someone, we can investigate the player and their in-game actions to determine whether they warrant an in-game suspension or ban.” IllFonic uses Amazon Elastic Container Registry (Amazon ECR)—a fully managed service that makes it easy for developers to store, manage, and deploy Docker container images—and Amazon Elastic Kubernetes Service (Amazon EKS) for container orchestration of backend microservices. “These AWS services act like little web servers and communicate with cache layers and the databases behind them,” says Jackson. “And in front, they communicate with all the players and game servers to provide information.”

IllFonic plans to be more proactive about finding and fixing in-game issues by diving into telemetry—storing and analyzing more information on what players are doing in games. “We want to gather information on what they’re playing, what they’re liking, and what they’re disliking,” explains Jackson. “We’ve been doing that currently to a small extent, but that’s something we definitely want to look at and expand in the long term.”

Migrating to Amazon GameLift has enabled IllFonic to provide a seamless gaming experience with low latency and better, faster matchmaking while reducing costs and work for its in-house staff. “Working on Amazon GameLift has made our small developing company look really good,” says Brungardt. “And it’s making our titles great to play. It definitely shows that a small company like us can compete and punch above its weight.”

To learn more, visit aws.amazon.com/gamelift/.