| Abstract |
Future defence infrastructure systems will require increased flexibility and agility to respond to changing application goals, external threats and complex environments. A key enabler for such agility is Software Defined Coalitions (SDC), where the network comprise multiple domains of resources owned by different defence units (partners) but dynamically joined together to form an infrastructure for communications and computation. Software Defined (SD) Slicing aims to enable agile and near-real-time provision and configuration of “slices” of the infrastructure resources for supporting future communications and computing applications. An SD slice, makes use of the allocated resources distributed across several domains to support a set of applications including distributed analytic services. We present a 3-level control architecture with a global SD-slice controller at the top, domain controllers (DC) in the middle, and dynamic, end-to-end flow controllers at the bottom. Associated with each DC is a domain inference engine whose function is to estimate availability of various resources in that domain. Based on the inferred resource availability in domains, the global controller determines the feasibility of supporting a new SD slice and if so, allocates resources to achieve/maintain the required performance of all slices. Based on the resources allocated by the global controller, the slice controller is responsible for sharing the resources across domains among data/processing flows to optimize resource utility. The end-to-end flow controllers then allocate resources to data flows or processing tasks according to dynamic conditions of resources for efficiency and robustness. |