Personal tools
You are here: Home Projects Service-Oriented Fog Computing with the Interconnected Mobile Edge of Things

Service-Oriented Fog Computing with the Interconnected Mobile Edge of Things

This project aims to address the integration-related challenges involved in applying Fog Computing in edge network of the Internet of Things (IoT) systems.


  • Software Defined Networking (SDN). SDN de-couples the management system from the hardware infrastructure in which the network administrator can control the network from the higher abstraction. The characteristics of SDN includes: dynamic, manageable, cost-effective, and adaptable. OpenFlow is a common approach to implement the SDN (Hu et al., 2014). Since it is expected that in near future, a large number of networked node (especially the routers and WiFi access points placed in many spots by Internet providers) will be interconnected to each other, the SDN can be applied in such environment to improve the overall performance of the network communication.
  • Cloudlet-based Virtual Machine (CloudletVM). CloudletVM (Satyanarayanan et al., 2009) is a concept that local businesses will ‘rent out’ their computer hardware resources for nearby mobile users for various computational and networking processes. The concept is based on installing Virtual Machine in box to provide the similar IaaS (Infrastructure as a Service)-based cloud capability from a cluster of machines in proximity-range. It can be realised by, for example, Open Edge Computing (OEC)’s Cloudlet solution (
  • Device-to-Device communication (D2D). Initial concept of D2D utilise existing cellular infrastructure (e.g. LTE-Advanced Networks) to assist mobile device communication in near-field. It can reduce the latency and improve the throughput (Doppler et al. 2009). Note that using cellular infrastructure to assist the communication among mobile nodes has been addressed by  Ericsson Research in 2008 (Srirama et al., 2009).
  • Mobile Edge Computing (MEC). MEC is an ongoing industrial standard founded by Huawei, IBM, Intel, Nokia Networks, NTT DOCOMO and Vodafone. The basic concept is to provide cloud computing capabilities within the near-field Radio Access Network (RAN) to mobile users. The characteristics of MEC includes: on-premises (each edge can be isolated), proximity, lower latency, location awareness and network context information. MEC is also based on the concept of utilising the cellular base stations to assist the fundamental infrastructure. The major difference between MEC and D2D is that in MEC, it is expected that the MEC server node can provide computational mechanisms such as real time video analytics in the near-field (Patel et al., 2014). MEC can be seen as a combination of the concept of CloudletVM and D2D.
  • Fog Computing (Fog). Fog Computing is the term created by CISCO (Bonomi et al, 2012). Fog shares many common characteristics with MEC but it has emphasis of its relationship with IoT and cloud. Similar to MEC in which Fog aims to bring the computational capabilities from cloud to near-field ground to reduce the latency (only the complex and high resource intensive tasks remain in cloud). Fog can be realised by routers (not necessary the cellular base stations) that have implemented with Fog node functionalities in which they can provide VM mechanism. One specific characteristic of Fog when comparing with MEC is the need of communication between edge nodes (Klas, 2015). As the technology is evolving, the near future IoT devices can be more capable in computational tasks and they can consume less energy, it can be foreseen that in the near future, various devices can participate in Fog (Vaquero and Rodero-Merino, 2014), which can provide even higher throughput, computational power, mobility, lower latency and scalability.

Applying Fog Computing in the edge network of the IoT system can highly improve the efficiency. However, numerous challenges need to be addressed:


  1. Discoverability—This is highly related to the mobility of the users of Fog. The process currently doing by the near-field Fog nodes may need to pass the result to another subnet in order to deliver the final result to the initiator since the initiator has been moved to another subnet. Further, while the initiator is moving, it needs to continuously discover the proximate environment in order to quick select the proper nodes for computational needs. It cannot be done by a central server, which is not applicable in Fog Computing. However, it is possible to utilise the super-peer network based P2P because in Fog Computing, there will be a router-centric mesh network exist to assist the environment.
  2. Limited computational power and storage of participants—Fog Computing requires a strategy to distribute the tasks in best performance and lowest latency. As the participants are heterogeneous in specification (hardware, network, storage and/or event platform/language support), a corresponding scheme/models for selecting nodes and distributing tasks is important. The scheme also need to consider the mobility.
  3. Management—there will be no central management party exist in Fog, its all about federated management. Clearly, a choreography approach is required. BPMS seems to be a promising approach to deal with this requirement. There will be also the need to distribute the management tasks to each subnet or each small cluster established by a number of participants, in which, the concept of BPM Everywhere is applied.
  4. Platforms—As Hyrax (Marinelli, 2009) has shown that MapReduce (Hadoop) is resource intensive and not feasible for resource constrained devices. Cloudlet (Satyanarayanan et al. 2009) is known that requires powerful machine to host the VM-in-box and Data Centre-in-box. Since Fog Computing will involve different type of devices, including less powerful devices (but within the same edge can be a powerful cluster). There is a stressed need of a novel sandboxed environment platform that can be hosted in different devices (including resource constrained devices) to enable them to participate in the Fog Computing. One possible solution is to support popular scripting language environment which leverages with the hardware components with the capability of handling the tasks given by the requesters. Also, there can be a promising approach to enable the process distribution based on service-oriented computing.
    The platform itself is not only provide the mechanism to perform “computation offloading” but also the need of data delivery between Fog Nodes and the Big Cloud. How data is moving between Fog Nodes and how tasks is performed and the result is delivered. These all need to be handled by the platform itself without the need of individual implementation of each node.
  5. Privacy and Reliability. The cloud-based IoT also has another issue in reliability. The cloud data centre failure can cause the whole system fail. However, it can be avoid by applying the Fog Computing concept. For example, the Netatmo camera service is always connected with its backend cloud server. The user who located in the same city needs to retrieve the data via the cloud server which locates in a far distant foreign country. With Fog, the data can be delivered to the user within the same country, even within the same city without go through all the way to oversea then comes back to the user. For example, suppose the Netatmo camera is connecting with Telia’s LAN in the user’s home, and the user (in office located at the another part of the city) is using Telia’s WAN. Within the same city’s network infrastructure, the data can already be delivered to the user within Telia’s network. The challenge is how to ensure the network provider does not see the data (privacy) but can deliver the data to the right user. The similar situation is applicable to other scenario such as real-time traffic sensing that uses multiple provider’s network. How does the data be delivered to the right person?





Chang, C., Srirama, S.N., Ling, S.,(2014), SPiCa: a social private cloud computing application framework, in Proceedings of the 13th International Conference on Mobile and Ubiquitous Multimedia (MUM '14). ACM, New York, NY, USA, 30-39.