A Novel Strategy to Achieve Bandwidth Cost Reduction and Load Balancing in a Cooperative Three-Layer Fog-Cloud Computing Environment

Abstract

Recently, IoT (Internet of Things) has been an attractive area of research to develop smart home, smart city environment. IoT sensors generate data stream continuously and majority of the IoT based applications are highly delay sensitive. The initially used cloud based IoT services suffers from higher delay and lack of efficient resources utilization. Fog computing is introduced to improve these problems by bringing cloud services near to edge in small scale and distributed nature. This work considers an integrated fogcloud environment to minimize resource cost and reduce delay to support real-time applications at a lower operational cost. We first present a cooperative three-layer fog-cloud computing environment, and propose a novel optimization model in this environment. This model has a composite objective function to minimize the bandwidth cost and provide load balancing. We consider balancing load in both links' bandwidth and servers' CPU processing capacity level. Simulation results show that our framework can minimize the bandwidth cost and balance the load by utilizing the cooperative environment effectively. We assign weight factors to each objective of the composite objective function to set the level of priority. When minimizing bandwidth cost gets higher priority, at first, the demand generated from the traffic generator sensors continues to be satisfied by the regional capacity of layer-1 fog. If the demand of a region goes beyond the capacity of that region, remaining demand is served by other regions layer-1 fog, then by layer-2 fog, and finally by the cloud. However, when load balancing is the priority, the demand is distributed among these resources to reduce delay. Link level load balancing can reduce the queueing delay of links while server level load balancing can decrease processing delay of servers in an overloaded situation. We further analyzed how the unit bandwidth cost, the average and maximum link utilization, the servers' resources utilization, and the average number of servers used vary with different levels of priority on different objectives. As a result, our optimization formulation allows tradeoff analysis in the cooperative three-layer fog-cloud computing environment. INDEX TERMS Fog Computing, IoT, Optimal Resource management, Load Balancing, Task Offloading I. INTRODUCTION Internet of Things (IoT) devices such as home voice controllers, smart TVs, smart locks in smart homes, road traffic and air quality monitors in smart cities, continuous glucose monitor, and fitness bands in healthcare are only a few examples, which are introduced in recent years. A huge growth in the number of IoT devices equipped with sensors is observed recently. These sensors collect the data from these devices. Later, the collected raw data is used to produce aggregated information and to take automated decision. It is projected that 26 billion IoT devices are to be installed by 2020 [1], and 75.44 billion by 2025 [2]. This is a fivefold increase from

Virtual Machine Resource Allocation Optimization in Cloud Computing Based on Multiobjective Genetic Algorithm

“…e linear programming method is adopted, with the minimum cost as the optimization objective, and the processing time of resource price, virtual machine quantity, and supplier's supply capacity as the optimization conditions. Formula (10) represents the optimization goal, namely, the overall cost; formula (11) indicates that the number of VMS obtained from the supplier in each usage phase does not exceed the number of VMS reserved. Equations ( 12)-( 15) indicate that the total amount of basic resources obtained from each supplier should not exceed the upper limit of the maximum resources that the supplier can provide.…”

“…The emergence of this technology has greatly improved the resources utilization. With the emergence of a shared resource pool, the number of servers in the cloud environment has reduced, thereby increasing the income of the cloud service provider [ 2 , 10 , 11 ]. And, resource allocation problems ultimately boils down to multiple application mapping relations between the virtual machine and the server, through reasonable allocation, adjusting the distribution of application virtual machines on the physical nodes, making full use of the service condition of the server idle resources, so as to reduce the amount of activation server, and thereby achieve the purpose of reducing energy consumption [ 12 ].…”

Online Reconfiguration of IoT Applications in the Fog: The Information-Coordination Trade-Off

“…Some works apply a similar approach to deal with the reconfiguration in the Fog. The authors in [7]- [11] study the provisioning problem, i.e., where to run the applications' components in the Fog infrastructure. In their proposals, the provisioning is modeled as an ILP (Integer Linear Programming) problem, considering the constraints in term of resources (e.g.…”

Differential Grey Wolf Load-Balanced Stochastic Bellman Deep Reinforced Resource Allocation in Fog Environment

“…In [7] an integrated fog-cloud environment was proposed with the goal of lowering resource allocation costs and minimizing latency. A cooperative three-layer fog-cloud computing environment was built for this purpose.…”

“…. , F n 􏼈 􏼉" Output: Latency-minimized optimal load balancing (1) Initialize "Rnd 1 ����→ ," "Rnd 2 ����→ ," "a" (2) Begin (3) For each dataset "DS" with task "T" and fog nodes "F" (4) Formulate cloud data centers with "n" numbers of servers "S" as in equations ( 1) and (2) (5) Model "VMs" processing different numbers of tasks "T" as in equations ( 3) and (4) (6) Update positions of wolves or the computing nodes by handling overutilization of fog as in equations ( 5) and ( 6) (7) Handle under-utilized fog detection using equations (9-11) (8) Estimate the final position vectors of the current individual as in equations (12-14) (9) Handle migration between virtual machines as in equation ( 15 As given in the above algorithm with the objective of minimizing the energy consumption during resource allocation between fog nodes by the load balancer, a novel Deep Reinforcement Learning model is designed. First, the state space representing the user requested task, time is obtained.…”