Instead of using the traditional approach of buying a fixed amount of servers that run continuously all month, with Amazon GameLift you only pay for the capacity you use. Amazon GameLift charges for instances by duration of use and bandwidth by quantity of data transferred. See a pricing example.

Typical multiplayer games use 50% of their peak server capacity on average. This means that half of the time their game servers have no active players. Amazon GameLift’s autoscaling feature can automatically turns off game servers when they are not needed and adds new game servers when more players arrive.

Amazon GameLift offers both On-Demand and Spot instances. With Spot instances you get the same high-performance hardware as On-Demand instances at savings up to 90%. You can achieve these savings while maintaining high game server availability with the use of FleetIQ, a feature that places new game sessions on game servers based on player latencies, instance prices, and Spot interruption rates. 

With Amazon GameLift Spot instance types you pay the Spot price that’s in effect at the beginning of each instance-hour for your running instance. If Spot price changes after you launch the instance, the new price is charged against the instance usage for the subsequent hour. Spot instance prices are set by Amazon Web Services and adjust gradually based on long-term trends in supply and demand for Spot instance capacity, but never exceed On-Demand prices. In the Amazon GameLift console, you can view three months of historical pricing and savings data for every instance type and region.

Instance Pricing: Linux | Windows

You are charged data transfer fees for traffic between your game servers and your game clients. Inbound data transfer is free. For outbound data transfer rates in AWS China (Beijing), visit the Amazon EC2 Pricing page.

This example uses a typical hourly player demand pattern for multiplayer games to illustrate Amazon GameLift pricing. In this example, we’ll assume that your game launches in Beijing and attracts 10,000 players. As shown below, the number of concurrent players follows a typical usage pattern.

In this example, your game has the fewest number of players in the morning (800 players at 7:00 am), a spike at lunch (4,350 players at 1:00 pm), and a daily peak in the evening (6,480 players at 10:00 pm).

Amazon GameLift offers instance types with different compute, memory, and networking capabilities. We’ll assume that you decide to host your game server using c4.xlarge instances running Linux. You've profiled your game server performance and found that each c4.xlarge instance can run 4 games with up to 10 players simultaneously.

Throughout the day, as your player demand shifts, you will need more or less capacity. For example, from 7:00 am to 8:00 am you will need 20 instances to meet player demand (800 concurrent players / 10 players per game / 4 games per instance = 20 instances). Below is your instance usage over a 24-hour period.

In this example, you decide to configure Amazon GameLift to automatically scale the number of instances up and down based on your current player demand. At 7:00 AM, you use 20 active instances, while at 10:00 PM, you use 162 active instances. Amazon GameLift only charges for the capacity you use. Therefore, you are only charged 20 instance hours from 7:00 AM to 8:00 AM, and 162 instance hours from 10:00 PM to 11:00 PM. The sum of all instances running each hour over the entire 24-hour period is 1,944 instance hours.

If you chose to use On-Demand instances, you would have been charged ¥ 2.773 per instance hour - the On-Demand price for an Amazon GameLift c4.xlarge instance running Linux in the AWS China (Beijing) region. Your total daily usage of 1,944 hours at ¥ 2.773 per hour sums to ¥ 5,391 per day, or ¥ 161,730 over a 30-day month.

If instead, you chose to use Spot instances, prices for your instances would have fluctuated throughout the day – but a good rule of thumb is that Spot instances running Linux are on average 70% cheaper than the same On-Demand instances. If you did see a 70% savings, your costs would be ¥ 1,617 per day, or ¥ 48,519 over a 30-day month.

If Amazon GameLift didn’t offer Auto-Scaling, but instead required you to pay for servers that run continuously all month, then your instance fees would have been ¥ 10,781 per day (¥ 2.773 * 162 instances * 24 hours) or ¥ 323,430 over a 30-day month. With Auto-Scaling, On-Demand instances would have saved you ¥ 161,700 per month, a 50% reduction over servers that run continuously all month, and Spot instances would have saved you ¥ 274,911 per month, an 85% reduction over servers that run continuously all month.

We'll assume that the total amount of monthly data transfer in the AWS China (Beijing) region from all of your game servers to your game clients is 49.82 terabytes per month. Based on standard Data Transfer rates in the AWS China (Beijing) region, your cost for 49.82 terabytes of data transfer is ¥ 46,482 per month.