As presented in the previous post of this blog, a central innovation of PrEstoCloud is to provide a state of the art management of resources. In other words, an entity capable of dynamically assigning computing tasks to resources while considering a complex inter-cloud infrastructure, heterogeneous cloud, and edge devices, and performance-driven application constraints.
The Autonomic Data-Intensive Application Manager (or ADIAM for short) is the central piece that connects the meta-management, communication, and infrastructure layers all together. This is the building block that allows resource provisioning and management to run big data applications.
As presented in the below figure, the PrEstoCloud Control Layer entry point is a topology and a set of placement constraints that describes the managed data-intensive application. This description is provided by the Meta-Management Layer. Following next, the Control Layer is responsible, in a first step, to compute and deploy the concrete infrastructure that becomes available to the managed application, and in subsequent steps, to dynamically adapt this infrastructure to fit specific execution requirements or conditions. The Communication Layer is essentially a message broker that provides all to all communication. Finally, once the computing infrastructure is deployed on chosen computing capable resources, the computation steps (application fragments) are dispatched and executed.
The Autonomic Data-Intensive Application Manager is the kernel of this logic in charge of orchestrating and automating data-intensive jobs life-cycle management.