Reserved capacity
Reserved capacity is a committed block of compute resources purchased at a discounted rate compared to on-demand pricing. You commit to a certain amount of capacity (measured in vCPUs or instances) and pay a lower hourly rate whether or not you use it all.
Reserved vs on-demand
- Reserved — Lower rate (e.g. $0.02/vCPU/hr), committed capacity, always available. You pay whether sessions are running or not.
- On-demand — Higher rate (e.g. $0.07/vCPU/hr), no commitment, pay only when sessions are active. Used for bursts above the baseline.
How studios use reserved capacity
Most multiplayer games have a predictable daily player pattern: low overnight, rising through the day, peak in the evening. Reserved capacity should cover the average daily load — the level your game consistently needs. On-demand handles the peaks.
For example: if your game averages 500 concurrent sessions and peaks at 2,000 during evening hours, you’d reserve capacity for 500 and let on-demand handle the remaining 1,500 during peak. This optimises cost while ensuring capacity is always available.
Reserved capacity and bare metal
In game server hosting, reserved capacity is typically backed by dedicated bare metal hardware. This gives you both the cost advantage of commitment pricing and the performance advantage of dedicated hardware — no noisy neighbours, consistent clock speeds, and included bandwidth.