Intent-Based Policy Management for Fog-Cloud Platforms

点击翻译按钮获取中文摘要

The architectural structure represented by cloud computing has worked very well for many applications. However, there are limitations that make it difficult to fully support certain types of applications. For example, an Internet of Things (IoT) application may require large amounts of data generated by sensors to be analyzed quickly in order to trigger actions in a timely fashion (e.g., traffic control, emergency response). Transferring data over the Internet to a cloud data center may incur significant delays that impact timely decision making which makes the use of current clouds unsuitable for applications with stringent real-time requirements. Furthermore, the transfer of large amounts of data over the network potentially could impact the costs associated with the network infrastructure needed to support the data transfer. To tackle these issues a new computing and networking paradigm, Fog Computing, has emerged. The main goal of fog computing is to distribute computing, storage, control, and networking services by bringing services to the network edge, where data is generated and where decisions will often need to be made in order to address stringent real-time requirements of applications. The result is a complex ecosystem of technologies, which we will refer to as the fog-cloud platform. Applications may require software components on both fog nodes and on clouds. The task of managing system health to ensure that applications are reliable and have the resources to satisfy their run-time requirements becomes more challenging due to the heterogeneity of the fog-cloud platform but also as the result of different owners. Currently, cloud data centers are owned by large businesses e.g., Amazon. It is anticipated that fog nodes will be owned by telecommunications companies e.g., Bell Canada. The resources needed are often dynamic over time. The implication is that management of the fog-cloud platform will require that management system used by the owners of the component systems are able to dynamically collaborate and coordinate the allocation of computing/network resources to satisfy application run-time requirements. Typically, there isn't a "one size fits all" management strategy. Policies were introduced to enable management systems to specify management strategies that are interpreted at run-time by management systems. This allows for changes in management strategies. Current approaches have worked well but as we move to fog-cloud platforms there is a need for innovative changes in the policy specification.  The overarching objective of my research program is the development of a policy-based management framework that supports management of the fog-cloud platform.  The results of the proposed research will have a direct impact on the management operations of the fog-cloud platforms enabling self-adaptive systems supporting more efficient resource use.