In the context of cloud computing, it refers to the ability of a cloud system to dynamically allocate and deallocate resources, such as compute power, storage, or networking, based on the current demands of the application or workload.
Two types of elasticity can be discovered depending on the time frame under study:
- Coarse-grain: when the entire infrastructure needs to scale over a period of minutes and hours.
- Fine-grain: when unpredictable users requests need to be assess at the second scale.
With the existing cloud infrastructure, both types of elasticity cannot be satisfied at the same time. VMs and containers are suitable for coarse-grain elasticity (slowly evolving loads) but not for fine-grain elasticity (high, unpredictable load bursts).
In the paper of Boxer, a new type of elasticity referred as ephemeral elasticity is proposed.