Elasticity

Book chapter, 2019

In data stream processing, elasticity refers to the ability of autonomously provisioning and decommissioning resources (e.g., threads or nodes) in order to match, at each point in time, the computational power needed to run a set of user-defined continuous queries. Elastic SPEs are able to dynamically scale up and out (or down and in, respectively) the resources employed to run the existing queries in order to match the needed computational power at each point in time. They can thus benefit of decoupled architectures such as cloud computing ones, and similar metrics for the performance evaluation can be defined (Ilyushkin et al. 2017). The variations in the needed computational power are due to the fluctuating nature of data streams (both in terms of their volume and their underlying data distribution) and the variable number of queries added and removed over time by users.