Scalable Extension of Volunteer Supported Fog Computing (VSFC)
Fog computing extends cloud services towards the network edge where end devices are located. This distributed computing approach is ideal especially for applications with real-time requirements. However, Fog has limited computing resources compared to cloud, hence it has to forward compute-intensive jobs to cloud which has the drawback of incurring higher latency. A practical and low-cost solution to this problem is the use of volunteer computing (VC) devices in the vicinity of Fog as compute nodes. Since these devices are in the vicinity of Fog, it may offload some of its compute tasks to nearby VC devices instead of cloud thus reducing latency.
In this thesis, we propose to evaluate this Volunteer Supported Fog Computing (VSFC) approach in terms of heterogeneity and scalability criteria, i.e., how feasible the approach is considering multiple VC devices of different processing capabilities and battery capacities. We also propose to work on scheduling algorithms which offload tasks from fog to VC devices and cloud keeping in view the constraints on latency, energy consumption and network usage. To this effect, we have further extended the iFogSim toolkit with provisions for VSFC to include multiple heterogenous VC devices and simple scheduling algorithms for placement of application modules on various available devices. We have considered two simulation setup scenarios in our work. The results show that the proposed VSFC multi-VC scheme outperforms the traditional delay-priority scheme and VSFC one-VC scheme by providing a 20% improvement in energy consumption and a striking 10 times reduction in network usage. In case of simulation setup 2, our scalability results show that for each addition of five VC devices we obtain a 50% improvement in energy consumption and 10 times improvement in network usage. Similarly, the heterogeneity results show that by allowing low power heterogenous VC devices we can reduce the static energy of VC devices as well.
Meeting ID: 951 6217 2994