Clanforge incorporates a handful of different strategies to ensure that you get the most out of your fleet with the lowest possible cost. Keeping machine capacity, whether cloud or bare metal has an associated cost depending on the following factors:
Machine type
Machine provider
Machine location
Machine specifications
Capacity types
Two types of machine capacity compose the total capacity of a fleet: bare metal machines and cloud machines.Bare metal machines often cost less than cloud machines for long-term use, so the reactive scaling system prioritizes bare metal capacity over cloud capacity. This is because cloud machine providers bill based on consumption (the more resources you consume, the more you pay) while bare metal providers charge a static price (you pay the same price despite the resources you consume).
Cost scores
The reactive scaling system decides which machine to employ by using a property on machines called a cost score. A cost score is a property value assigned to machines at the region location level that helps the reactive scaling system determine the most cost-effective machine for an allocation. The less cost efficient the machine, the higher the cost score. bare metal machines always have a lower cost score than cloud machines.
Best available server
The best available server is an available game server on the most cost-effective machine that meets the requirements for an allocation request. While the cost score of a machine matters in determining the best available server, there are other considerations, such as the location, state, and connection quality of the machine. In allocation-based fleets, the reactive scaler automatically determines the best available server per allocation request. In reservation-based fleets, there is no automatic selection.
Capacity limits
It’s possible to configure fleets and fleet regions with hard limits on machine capacity to avoid the additional costs associated with scaling past a certain number of machines. The maximum value in the fleet and fleet region configurations determines this capacity limit.
Time to live settings
The time to live (TTL) values in your fleet and fleet region configurations impact the overall cost of keeping your fleet by preventing unused machines from staying online for too long, which would lead to unnecessary costs. Refer to fleet scaling settings.
Server defragmentation
To optimize the cost of keeping your fleet, the reactive scaler aims to consolidate game servers into as few machines as possible. It does so by trying to allocate all the game servers on a machine before allocating the game servers on another machine of equal cost-efficiency. By doing this, the scaler avoids situations where a fleet has multiple machines with unused game servers. Refer to server fragmentation.