Dynamic Resource Allocation and Pricing for Fog-Cloud Networking as a Service

点击翻译按钮获取中文摘要

The Fog computing (comprising the Edge computing) concepts are gaining a significant momentum since they can provide low service latency combined with geographically distributed computing, independent of the user locations. These ingredients can significantly empower Tactile Internet, IoT and multi-user applications. To achieve these objectives, a well-developed and economically viable infrastructure for the Fog-Cloud computing is needed. In this proposal, we address some important issues related to this goal. In particular, we believe that the applications should have access to multi-node and multi-layer Fog-Cloud computation resources, enabling distributed computations, where the networking aspect is of great importance. The considered computation nodes can belong to different owners and can include edge nodes, local data centers, and distant data centers. Managing resource allocation in such a distributed set of computing and networking resources can be too complicated for users and applications. To address this issue we propose a service that we name Fog-Cloud Networking as a Service (FCNaaS) where the FCNaaS operator (FCNO) combines the role of an overlay network and a broker. The overlay network part manages the networking and computing resource allocation to a given demand while the broker part facilitates long-time-scale interaction between the users, applications and networking-computing resource owners that involves resource reservations for FCNO and pricing. The research will be focused on three areas related to FCNaaS: 1) Resource management architecture for Fog-Cloud Networking, 2) Dynamic resource allocation and related economic framework, and 3) Pricing and resource reservations. In particular, we will develop AI based resource allocation models within an economic framework that takes into account heterogeneous environment with several resource and application providers. Also, using a game theoretical framework we will develop models for pricing that give incentives to each player to participate in the market in an economically efficient and balanced way that provides benefits to all players. The research results will be useful to Canadian application providers, network operators, and computing node owners for optimizing their operations and to grow their market shares. The developed knowledge and models will be also useful to the government regulatory bodies for providing social wellness of the offered services. Also, the training of HQP is an integral part of the proposed research program and the students will acquire important skills in the domain that is becoming of high importance to several Canadian industries.