Fleet scaling settings#

Fleet scaling settings exist in two locations: at the fleet level and the fleet region level. At both these levels, you can specify most scaling setting values.

At the fleet region level, in particular, you can:

• Override values specified at the fleet level
• Control the preferred settings for the reactive scaling system

The hierarchy of fleet scaling settings allows you to create default settings for a fleet and control region-specific settings by selectively overriding settings per fleet region.

Each fleet has a collection of scaling settings that, by default, apply to all fleet regions within the fleet. If a fleet region demands a unique set of scaling settings, you can explicitly specify those settings for that fleet region. This hierarchy allows you to customize the scaling behavior of a fleet for each region. Although you can override most scaling settings per fleet region, you can only specify the Allocation spread value per fleet.

The topics below explain the purpose of each fleet scaling setting.

Minimum available servers#

The minimum available servers value determines the minimum number of game servers that the reactive scaling system keeps online and available for new players to join. Keeping a buffer of available game servers reduces the likelihood that players have to wait for the reactive scaling system to start up a cloud machine before they can join a game session. The minimum available servers value has the most influence over the reactive scaling system’s scaling decisions. Refer to Available servers buffer.

There are no limits on the number of available servers you can have in your fleet unless you specify a Maximum servers value. The higher the minimum available servers value, the faster your fleet or fleet region responds to sudden increases in players. Higher values also incur a higher maintenance cost.

Game Server Hosting advises specifying the minimum available servers at the fleet region level because the ideal value depends on the size of the region and the expected maximum rate of allocation requests in the region.

Minimum servers#

The minimum servers value determines the absolute minimum number game servers that the reactive scaling system keeps in a fleet or fleet region at any time. This value prevents the reactive scaling system from removing excess capacity. It also ensures the reactive scaling system keeps an additional buffer of offline capacity that the reactive scaler can bring online quickly.

Minimum standby servers#

The minimum standby servers value dictates the number of game servers within a fleet or fleet region that are in an offline state. Game servers in a standby state are distinct from available game servers in that available game servers are online, but standby game servers aren't. Standby game servers serve as a form of warm capacity that Game Server Hosting can use to host game servers in case of a sharp influx in connected users (CCU). This setting is optional.

Maximum servers#

The maximum servers value is the total maximum number of game server instances that can exist at a time. You can change this value to control costs associated with scaling into the cloud. The maximum number of servers sets an upper limit on the total number of game server instances in your fleet. If you reach the maximum number of servers, your fleet stops scaling.

Allocation timeout#

The allocation timeout value determines the time the reactive scaling system waits before force-deallocating allocated game servers, even if players are on the server. This value serves as a hard limit on the time an allocation can exist.

Allocation spread#

The allocation spread value controls how the reactive scaling system spreads allocations across machines in a fleet or fleet region. The value represents the maximum number of machines the reactive scaler spreads allocations across. The reactive scaler uses a random value between one and the specified allocation spread value.

Specifying an allocation spread value that's higher than the number of bare metal machines can't result in additional scaling, but might cause the reactive scaling system to keep machines online longer than necessary. Refer to Allocation spread.

Time to live (TTL) values#

Time to live values, called TTL values for short, are configuration settings specified at the fleet region level that prevent the reactive scaling system from scaling machines up or down too quickly. Without the TTL settings, the reactive scaling system might add or remove machine capacity too quickly, which can cause increased cost and reduced performance.

There are TTL settings that control the time that machines exist in states before the reactive scaling system takes action, such as:

• The Shutdown TTL, which is the time a machine exists in an unused state before the reactive scaling system shuts it down
• The Delete TTL, which is the time a cloud machine exists in a shutdown state before the reactive scaling system deletes it
• The Disabled delete TTL, which is the time a cloud machine exists in a disabled state before the reactive scaling system shuts it down

Shutdown TTL#

The shutdown time to live (TTL) determines the time the reactive scaling system waits before shutting down a cloud machine with no allocated game server instances. A cloud machine must remain inactive for at least the shutdown TTL before the reactive scaling system shuts it down. The shutdown TTL applies to each cloud machine individually after all the game servers on it are free of players if the parent fleet or fleet region has enough capacity to meet the Minimum available servers value. The higher the shutdown TTL value, the longer it takes for your fleet or fleet region to scale down.

Delete TTL#

The delete time to live (TTL) determines the time the reactive scaling system waits before deleting a cloud machine after shutting it down. The delete TTL applies to each cloud machine individually as soon as it's shutdown. When a cloud machine reaches the delete TTL, the reactive scaling system deletes the instance from the providing cloud provider. The higher the delete TTL, the longer it takes for your fleet or fleet region to scale down.

Disabled delete TTL#

The disabled delete TTL (time to live) value determines the time the reactive scaling system waits before deleting a disabled cloud machine. It's like the Delete TTL value, except it only applies to disabled cloud machines.

Start on provision#

The start on provision setting control whether the reactive scaler automatically started all the game servers on machines after bringing the machines online.

When the start on provision setting is enabled, the reactive scaler starts all the game servers on machines after bringing the machines online. When the start on provision setting is disabled, the reactive scaler doesn't start the game servers on machines automatically; instead individual game servers are started as they're allocated.