Towards Distributed Data Management in Fog Computing

Abstract

In the emerging area of the Internet of Things (IoT), the exponential growth of the number of smart devices leads to a growing need for efficient data storage mechanisms. Cloud Computing was an efficient solution so far to store and manipulate such huge amount of data. However, in the next years it is expected that Cloud Computing will be unable to handle the huge amount of the IoT devices efficiently due to bandwidth limitations. An arising technology which promises to overwhelm many drawbacks in large-scale networks in IoT is Fog Computing. Fog Computing provides high-quality Cloud services in the physical proximity of mobile users. Computational power and storage capacity could be offered from the Fog, with low latency and high bandwidth. This survey discusses the main features of Fog Computing, introduces representative simulators and tools, highlights the benefits of Fog Computing in line with the applications of large-scale IoT networks, and identifies various aspects of issues we may encounter when designing and implementing social IoT systems in the context of the Fog Computing paradigm. The rationale behind this work lies in the data storage discussion which is performed by taking into account the importance of storage capabilities in modern Fog Computing systems. In addition, we provide a comprehensive comparison among previously developed distributed data storage systems which consist of a promising solution for data storage allocation in Fog Computing.

How to place your apps in the fog: State of the art and open challenges

“…Fog computing architectures substantially differ from cloud architectures. In particular, according to recent literature: –Fog nodes feature limited and very heterogeneous resources, while data center nodes feature high (and virtually unbounded) computational, storage and power capabilities. –Fog devices are highly geographically distributed—often mobile—and possibly span wide large‐scale networks reaching closer to (human and machine) end users, while cloud data centers are located in few geographic locations all over the world and connect directly to fibre backbones. –End‐to‐end latency between cloud nodes within data centers is usually negligible and bandwidth availability is guaranteed via redundant links, while in fog domains, network QoS between nodes can largely vary due to the presence of a plethora of different (wired or wireless) communication and Internet access technologies. –Fog nodes are owned and managed by various service providers (from end users to Internet service providers to cloud operators) and might also opportunistically include available edge devices (eg, crowd computing, ad hoc networks) whereas the largest cloud data centers are in the hands of a few big players. …”

“…Fog computing architectures substantially differ from cloud architectures. In particular, according to recent literature [12][13][14][15][16] : -Fog nodes feature limited and very heterogeneous resources, while data center nodes feature high (and virtually unbounded) computational, storage and power capabilities. -Fog devices are highly geographically distributed-often mobile-and possibly span wide large-scale networks reaching closer to (human and machine) end users, while cloud data centers are located in few geographic locations all over the world and connect directly to fibre backbones.…”

Support Mobile Fog Computing Test in piFogBedII

“…But the location of end devices is static and cannot be updated; on the other hand, it is limited to the discrete event simulators (DES) and has very poor scalability because of the characteristics of CloudSim. Myifogsim [17] extends iFogSim to support mobility by migrating virtual machines between cloudlets. In [18], Naas extends iFogSim to implement data deployment strategy in fog and IoT.…”

QoS-Aware Dispersed Dynamic Mapping of Virtual Sensors in Sensor-Cloud

“…Thus, sensorcloud emerges as a highly scalable and easily accessible system, which is capable of serving a huge number of users [1], [2]. Essentially, the Sensor-Cloud Infrastructure is built on the concept of virtualisation of hardware resources of cloud computing [3], thereby enabling the same physical sensor nodes to be used for serving multiple end-user applications simultaneously. Similar to other cloud-based infrastructures, a centralized Sensor Cloud Service Provider (SCSP) obtains physical wireless sensor nodes from their respective sensor-owners.…”