Deep Neural Networks (DNNs) have served as a catalyst in introducing a plethora of next-generation services in the era of Internet of Things (IoT), thanks to the availability of massive amounts of data collected by the objects on the edge. Currently, DNN models are used to deliver many Artificial Intelligence (AI) services that include image and natural language processing, speech recognition, and robotics. Accordingly, such services utilize various DNN models that make it computationally intensive for deployment on the edge devices alone. Thus, most AI models are offloaded to distant cloud data centers (CDCs), which tend to consolidate large amounts of computing and storage resources into one or more CDCs. Deploying services in the CDC will inevitably lead to excessive latencies and overall increase in power consumption. Instead, fog computing allows for cloud services to be extended to the edge of the network, which allows for data processing to be performed closer to the end-user device. However, different from cloud data centers, fog nodes have limited computational power and are highly distributed in the network. In this paper, using Mixed Integer Linear Programming (MILP), we formulate the placement of DNN inference models, which is abstracted as a network embedding problem in a Cloud Fog Network (CFN) architecture, where power savings are introduced through trade-offs between processing and networking. We study the performance of the CFN architecture by comparing the energy savings when compared to the baseline approach which is the CDC. CCS CONCEPTS• Computing methodologies → Machine learning; • Networks → Network architectures.

In the near future, IoT based application services are anticipated to collect massive amounts of data on which complex and diverse tasks are expected to be performed. Machine learning algorithms such as Artificial Neural Networks (ANN) are increasingly used in smart environments to predict the output for a given problem based on a set of tuning parameters as the input. To this end, we present an energy efficient neural network (EE-NN) service embedding framework for IoT based smart homes. The developed framework considers the idea of Service Oriented Architecture (SOA) to provide service abstraction for multiple complex modules of a NN which can be used by a higher application layer. We utilize Mixed Integer Linear Programming (MILP) to formulate the embedding problem to minimize the total power consumption of networking and processing simultaneously. The results of the MILP model show that our optimized NN can save up to 86% by embedding processing modules in IoT devices and up to 72% in fog nodes due to the limited capacity of IoT devices.

Massive amounts of data are expected to be generated by the billions of objects that form the Internet of Things (IoT). A variety of automated services such as monitoring will largely depend on the use of different Machine Learning (ML) algorithms. Traditionally, ML models are processed by centralized cloud data centers, where IoT readings are offloaded to the cloud via multiple networking hops in the access, metro, and core layers. This approach will inevitably lead to excessive networking power consumptions as well as Quality-of-Service (QoS) degradation such as increased latency. Instead, in this paper, we propose a distributed ML approach where the processing can take place in intermediary devices such as IoT nodes and fog servers in addition to the cloud. We abstract the ML models into Virtual Service Requests (VSRs) to represent multiple interconnected layers of a Deep Neural Network (DNN). Using Mixed Integer Linear Programming (MILP), we design an optimization model that allocates the layers of a DNN in a Cloud/Fog Network (CFN) in an energy efficient way. We evaluate the impact of DNN input distribution on the performance of the CFN and compare the energy efficiency of this approach to the baseline where all layers of DNNs are processed in the centralized Cloud Data Center (CDC).

No abstract