Advertisement

Metaheuristic Optimization Technique for Load Balancing in Cloud-Fog Environment Integrated with Smart Grid

Conference paper
Part of the Lecture Notes on Data Engineering and Communications Technologies book series (LNDECT, volume 22)

Abstract

Energy Management System (EMS) is necessary to maintain the balance between electricity consumption and distribution. The huge number of Internet of Things (IoTs) generate the complex amount of data which causes latency in the processing time of Smart Grid (SG). Cloud computing provides its platform for high speed processing. The SG and cloud computing integration helps to improve the EMS for the consumers and utility. In this paper, in order to enhance the speed of cloud computing processing edge computing is introduced, it is also known as fog computing. Fog computing is a complement of cloud computing performing on behalf of cloud. In the proposed scenario numbers of clusters are taken from all over the world based on six regions. Each region contains two clusters and two fogs. Fogs are assigned using the service broker policies to process the request. Each fog contains four to nine Virtual Machines (VMs). For the allocation of VMs Round Robin (RR), throttle and Ant Colony Optimization (ACO) algorithms are used. The paper is based on comparative discussion of these load balancing algorithms.

Keywords

Microgrid Cloud computing Fog computing Renewable energy sources Demand side management Ant colony optimization 

1 Introduction

An energy management system (EMS) is necessary to control the electricity consumption and generation. Platform for the EMS should be economical and affordable for the ordinary electricity consumer. Microgrid (MG) now made the Power Grids (PGs) more efficient by its unique features like distributed generation and energy storage, MG can connect with PG or can operate independently [1, 2]. The generated energy from MG need to be distributed among consumers using Smart Grid (SG) to maintain the balance between energy demand and supply curve in on-peak and off-peak hours [3, 4, 5]. Smart meters are used to access the information of Internet of Things (IoT) devices. IoT includes the number of devices that can sense, act and compute [6]. SG is completely automated infrastructure which controls the energy requirements of these devices in the economical way with the responsibility of rapid response [1, 7].

The SG is modern electrical grid, it supports two way communications, which helps to gather the information from different IoT devices using Information and Communication Technology (ICT). The energy generation is managed, devices are monitored and energy is distributed among them by using that gathered information. SG improves reliability, sustainability and efficiency of smart meters. It manages the operations, cost and reduce the peak load for utilities and transmits electricity in the efficient way in economical cost [8].

When billions of Internet of Thing (IoT) devices are in operation at that time the calculation for these devices become burden for the SG. To overcome this issue cloud computing introduced as a centralized computing, integrated with IoT. Cloud computing provides different hardware and software base services over the network. Super computer processing can be achieved by using the cloud computing [9]. It provides its platform for different computational service i.e. user can execute the code or can perform any type of processing. Cloud provides software as a service facility to the users, any one can view his/her profile through browser. User can use Infrastructure of a cloud as a service e.g. server can be used as storage [10].

When the number of devices increased in IoT the processing smart meters may take the extra computing time and response time will be increased due to computing latency [1]. Fog computing introduced to cover this limitation. Fog Computing is laying between IoT devices and cloud data center. Fog provides the storage, computation and networking services between devices and cloud computing. Day by day the number of IoT devices are increasing, the traditional cloud system is not designed for the type of data which IoT devices produce [4]. Other unique features of fog includes; mobility, heterogeneity, location awareness and real time response. The reason for these advantages is nearer to the end user, instead of remote data center.

In this paper, interaction between the SG and the emerging technology fog computing is considered and proposed the Ant Colony Optimization (ACO) algorithm for the load balancing of VMs lying in the fog architecture. At the end results are compared with Round Robin (RR) and throttle algorithms.

1.1 Motivation

For the implementation of EMS, a platform is necessary for the interaction of devices [1]. Author proposed a Fog based EMS, where fog is introduced as a platform for the connectivity of smart devices, which is based on low cost and low power devices. Fog computing provides the facility of separate storage of private and public data because of its distributed nature [11]. The main reason for employing cloud computing is its cheap rates. In [10], considered the problem of charging and discharging of Electrical Vehicles at Supply station, Fogs are working locally under the central cloud. Two different scheduling algorithms are proposed for minimizing the waiting time of vehicles. Author in [12] proposed a cloud-fog based environment and used Particle Swarm Optimization (PSO) and RR for the load balancing of VMs, because burden on a particular VM can lead to high latency, so balancing of load is necessary for effective response.

1.2 Contribution

In this paper, proposed the cloud-fog based infrastructure, integrated with SG to get benefits of cloud computing. Following are the contributions of the paper.

  • Six regions are considered from all over the world for managing the electricity consumption using fog layer as a central layer of architecture. Fog layer stores the information of consumer layer at local region level.

  • Different numbers of VMs are allocated to the fogs for the simulation purpose ranges from four to nine.

  • For the allocation of VMs to manage response time, processing time and cost, Ant Colony Optimization (ACO) technique is used.

  • ACO technique is farther compared with RR and throttle algorithm. It has been observed that proposed algorithm performed better than RR in all mentioned aspects and closer to throttle.

The rest of the paper is organized as: Sect. 2 related work, Sect. 3 introduces system model, Sect. 4 based on the simulation and result discussion and at the end conclusion of the paper is drawn.

2 Related Work

To balance the load on cloud and fog multiple algorithms are proposed by the researchers. These algorithms manage the requests of clients and schedule their tasks. The load balancing algorithm assigns task to each VM to utilize all the resources and to avoid the unbalancing between the VM’s load. The common algorithms used for the resource allocation are, Round Robin (RR), Shortest Job First (SJF), throttle, Artificial Bee Colony (ABC), Genetic Algorithms (GA) etc.

The authors in [10] formulated the problems of Electrical Vehicles (EVs) charging and discharging. Interaction between the EVs, Electrical Vehicle Power Supply Station (EVPSS) and cloud computing is defined. Scheduling is performed by using two priority assignment algorithms based on calendar and non calendar vehicles. Fog computing platform is used to manage the computation of devices in [1] to get the high response rate with low latency.

In [13] authors considered the efficient management system of mobile phones. Mobile phones always retrieve the location of data from the cloud. However, retrieving always data from the remote cloud, it will increase the response time so the edge computing concept is utilized to reduce latency and get faster response. Approximate approach is utilized to formulate Primal Problem (PP) into three Sub Problems (SPs). Particle Swarm Optimization (PSO) with simulated annealing algorithm is used for the resource allocation in [14] with the new service broker policy to allocate the data center by least estimated response time. Finally proposed technique compared with RR, throttle and Cuckoo search.

Shortest Job First (SJF) is proposed in [15] to tackle the load. The algorithm is implemented in two different scenarios in the paper. In first scenario one data center is allocated to twenty five VMs and in second two data centers are allocated to fifty VMs. Second scenario produces better results than first one. This proposed technique is farther also compared with RR and Equally Spread Current Execution (ESCE) and examined that SJF produced better result than others in the proposed scenario.

In [16] author proposed real-time dynamic pricing model and scheduling the users request on real time environment. Considered problem is of electric vehicles charging and discharging to reduce the peak load and manage requests efficiently. Software define networking technology based decentralized cloud computing architecture is used in the proposed approach. Real load of Toronto electricity is used for the simulations. For reducing load Peak to Average Ratio (PAR) Monte Carlo method is used in [17]. By reducing load from PAR reduces the cost of electricity.

[18] focuses on the novel cost oriented optimization model to allocate the cloud computing resources to the cloud based Information and Communication Technology (ICT) infrastructure. Resources are allocated in a cost efficient and flexible way using the modified priority list algorithms. To demonstrate the effectiveness of the proposed model on the operation cost reduction, comprehensive numerical studies are included.

Authors in [12] used RR, throttle and Particle Swarm Optimization (PSO) for the load balancing of VMs. For the experimental purpose two clusters are considered in two regions. Each cluster linked individually with one fog. Overall performance of PSO is better than the throttle and RR. In [7] three service broker policies; service proximity policy, optimize response time, dynamically reconfigure with load are compared with newly purposed service broker policy. Each cluster linked with two fogs, service broker policy decides the host of requested service from these two fogs. Two load balancing algorithm RR and throttle also used in this paper. Newly proposed service broker policy produces better results with throttle rather than RR.

3 System Model

In the proposed model, 6 regions are considered from all over the world. Each region includes two clusters and each cluster contains four to five buildings. The model is based on three layers, i.e., consumer layer, fog layer and cloud Layer. In a single region, we have two fogs, each fog is linked with two clusters via controller, as shown in Fig. 1. We consider that in each building, we have seven to ten homes. Each home has its own renewable energy resources and battery to save extra energy. When the demand of electricity is higher than the produced energy, then the storage battery will be utilized in order to fulfill the electricity demand. In case, battery does not have enough energy storage as compared to user’s demand, then the request is send to that fog, by smart meter of a home. Fog transfers the request towards the SG. Numbers of appliances are here to consume electricity.

The fogs are used to allocate the resources in an efficient way. Homes are connected with fog through smart meters. Fog contains different VMs to provide fast response to the consumers. Numbers of different applications are running on the VMs to process the requests of clients.
Fig. 1.

Proposed system model

Cloud layer is on the top of fog layer. Fog provides the facilities of cloud, faster than the cloud because cloud is far away than fog from the user. For the permanent storage of data, fog sends it towards the cloud because it can only save data temporarily. The main reason of using cloud-fog based environment on Smart Grid is for computational purpose. Different load balancing Algorithms are used for the balancing of the load on VMs.

3.1 Problem Formulation

In proposed system model we considered M fogs in each region, each fog contain set of VMs as \(\text {VM}=\{vm_1, vm_2, vm_3,\dots ,vm_m\}\). Set of VMs is denoted by VM and there are total m numbers of VMs. These VMs deal different tasks which are assigned by the load balancing Algorithm and requested by the users. Set of tasks is represented as \(\text {T}=\{t_1,t_2,t_3,\dots ,t_n\}\). There are n numbers of total tasks which will be served by different VMs.

The VMs are working in parallel order, each VM have the same capacity to deal with different tasks. These VMs do not share resources with each other. Each VM is assigned by evaluating different performance metrics: response time, processing time, cost of VM, requests per hour. These parameters are formulated using linear programming [19, 20].

Processing time of each VM is indicated using PT. To check task is assigned or not, binary representation is given in Eq. 1. where 0 is representing there is no task assigned to VM. 1 is used to indicate that task is assigned to VM.
$$\begin{aligned} t_{ij} = {\left\{ \begin{array}{ll} 1; &{} \text {If VM is assigned }\\ 0; &{} \text {else } \end{array}\right. } \end{aligned}$$
(1)
Minimizing PTt is the objective of this proposed work. PTt represent the total processing time of VM, tij indicates i requests assigned to j servers in specific time period. PTij indicates the processing time of each request i on the server j.
$$\begin{aligned} \text {Minimize } PT_{t}=\sum _{i=1}^{m} \sum _{j=1}^{n}(PT_{ij}*t_{ij}) \end{aligned}$$
$$\begin{aligned} \text {Subject to:} \sum _{i=1}^{m} t_{ij}=1, j=1,2,...,n \end{aligned}$$
(2)
$$\begin{aligned} t_{ij}=0\text { or }1. \end{aligned}$$
The total Response time of VM is calculated by using Eq. 3. CTx is the completion time of task. The cost of the system is calculated summing up the cost of VMs, MGs and data transfer cost. VM cost is computed by multiple instructions per second and the required task size. Makesspan term is used for describing the total execution time of task. The total cost is computed using the Eq. 4.
$$\begin{aligned} RT=\frac{\sum VMs(CT_x)}{makesspan*NumofVMs} \end{aligned}$$
(3)
$$\begin{aligned} C_t=C_{DT}+C_{VM}+C_{MG} \end{aligned}$$
(4)
Ct representing the total cost, CDT representing the data transfer cost, CVM representing the cost of VM and CMG representing the cost of microgids.

3.2 Algorithm

The proposed algorithm is ACO, it is used to find the best path from the wide range by selecting optimal path. Ants in real world wander randomly for finding food, after finding food by lying down their pheromone trail return to their colony.

The fellow ants, follow the same trail for food by discovering the shortest path.

4 Simulation and Discussion

As mentioned in purposed model, we have considered six regions in the scenario. Regions are representing continents of the world. Total twelve fogs are working under a central cloud. Each Region contains two fogs and two clusters of consumers. These two fogs and two clusters are meshed with each other, in each region. By using them we have simulated our result.

4.1 Simulation Setup

Java platform is used for simulations with Netbeans. By utilizing the ‘Cloud Analyst tool’ results of ACO are computed and also compared with two other load balancing techniques, i.e., throttle and RR. For the selection of fog three services broker policies are available in ‘Cloud analyst’. For plotting Graphs, Matlab is used. Machine for performing the simulation is utilized with processor ‘Intel Core i3’, RAM 6 GB and storage 500 GB.

4.2 Discussion

The performance measures of proposed scheme are: data center request servicing time, response time and cost.
Fig. 2.

Response time by region

Figure 2 shows the response time by region, which is provided by fogs to their dependent clusters. RR and Throttle result also shown in Fig with ACO. This graph is plotted using average response time of these three techniques. As shown in Table 1 the overall minimum, maximum and average response time of RR is 35.63, 67.72 and 51.05 respectively, with throttled and ACO response time is 35.63, 67.62, 51.00 and 35.72, 66.62 and 51.01 respectively.
Table 1.

Overall response time

Algorithm

Avg processing time (ms)

Min processing time (ms)

Max processing time (ms)

RR

51.05

35.63

67.62

ACO

51.01

35.72

66.62

Throttle

51.00

35.63

67.62

Fig. 3.

DC request processing time

Table 2.

Overall processing time

Algorithm

Avg processing time (ms)

Min processing time (ms)

Max processing time (ms)

RR

1.35

0.01

7.57

ACO

1.31

0.01

10.62

Throttle

1.30

0.01

5.07

Request processing time is depicted in Fig. 3 with RR, throttle and ACO. This graph is also plotted using average processing time of these techniques. In Table 2 we can see that the overall minimum, maximum and average processing time of RR is 0.01, 7.57 and 1.35 respectively, however, with throttle and ACO processing time is 0.01, 5.07, 1.30 and 0.01, 10.62 and 1.31 respectively. We can observe by considering average processing time that RR is high as compare to throttle and ACO, then ACO has less and throttle has least.

Figure 4 shows the cost comparison, it is plotted using total cost of RR, throttle and ACO. Table 3 showing the total VM cost, data transfer cost and total cost. Firstly of RR is 204.01, 160.49 and 364.50 respectively, secondly throttled 204.01, 160.49, 364.50 and lastly ACO cost is 204.01, 160.49 and 364.50 which same. From the Results we can conclude that the proposed Algorithms ACO is overall performing better than RR and nearer to throttle.
Fig. 4.

Total cost

Table 3.

Overall cost

Algorithm

VM cost

Data transfer cost

Total cost

RR

204.01

160.49

364.50

ACO

204.01

160.49

364.50

Throttle

204.01

160.49

364.50

5 Conclusion

In this paper, a cloud-fog based architecture is proposed based on three layers; cloud layer, fog layer and consumer’s layer. MGs are connected with fogs, user cannot communicate directly with it. Users put their request on fog for the energy requirement. These requests are assigned to fogs by using three different service broker policies. Fog layer compute the energy, consumed by the consumers and process their requests using VMs. Load of VMs is balanced by using RR and throttle and ACO load balancing algorithm. Simulation is performed using Java platform in NetBeans. Simulation results shows that the proposed algorithm ACO performed better than the RR.

References

  1. 1.
    Al Faruque, M.A., Vatanparva, K.: Energy Management-as-a-Service Over Fog Computing PlatformGoogle Scholar
  2. 2.
    Yoldaş, Y., Önen, A., Muyeen, S.M., Vasilakos, A.V., Alan, İ.: Enhancing Smart Grid With Microgrids: Challenges And OpportunitiesGoogle Scholar
  3. 3.
    Kumar, N., Zeadally, S., Misra, S.C.: Mobile cloud networking for efficient energy management in smart grid cyber-physical systems. IEEE Wirel. Commun. 23(5), 100–108 (2016)CrossRefGoogle Scholar
  4. 4.
    Kumar, N., Vasilakos, A.V., Rodrigues, J.J.P.C.: A Multi-Tenant Cloud-Based DC Nano Grid for Self-Sustained Smart Buildings in Smart CitiesGoogle Scholar
  5. 5.
    Mallada, E., Zhao, C., Low, S.: Optimal load-side control for frequency regulation in smart grids. IEEE Trans. Autom. Control 62(12), 6294–6309 (2017)MathSciNetCrossRefGoogle Scholar
  6. 6.
    Risteska Stojkoska, B.L., Trivodaliev, K.V.: A Review of Internet of Things for Smart Home: Challenges And SolutionsGoogle Scholar
  7. 7.
    Fatima, I., Javaid, N., Iqbal, M.N., Shafi, I., Anjum, A., Memon, U.: Integration of cloud and fog based environment for effective resource distribution in smart buildings. In: 14th IEEE International Wireless Communications and Mobile Computing ConferenceGoogle Scholar
  8. 8.
    Yaghmaee M.H., Moghaddassian M., Garcia A.L.: Power consumption scheduling for future connected smart homes using bi-level cost-wise optimization approach. In: Leon-Garcia, A., et al. (eds.) Smart City 360\(^\circ \), SmartCity 360 2016, SmartCity 360 2015 (2016)Google Scholar
  9. 9.
    Rajeshirke, N., Sawant, R., Sawant, S., Shaikh, H.: Load Balancing In Cloud Computing (2017)Google Scholar
  10. 10.
    Chekired, D.A., Khoukhi, L.: Smart Grid Solution for Charging and Discharging Services Based on Cloud Computing SchedulingGoogle Scholar
  11. 11.
    Okay, F.Y., Ozdemir, S.: A Fog Computing Based Smart Grid ModeGoogle Scholar
  12. 12.
    Zahoor, S., Javaid, N., Khan, A., Ruqia, B., Muhammad, F.J., Guizani, M.: A Cloud-Fog-Based Smart Grid Model for Efficient Resource UtilizationGoogle Scholar
  13. 13.
    Deng, R., Lu, R., Lai, C., Luan, T.H., Liang, H.: Optimal Workload Allocation in Fog-Cloud Computing Toward Balanced Delay and Power ConsumptionGoogle Scholar
  14. 14.
    Yasmeen, A., Javaid, N.: Exploiting Load Balancing Algorithms for Resource Allocation in Cloud and Fog Based InfrastructuresGoogle Scholar
  15. 15.
    Javaid, S., Javaid, N., Tayyaba, S.K., Sattar, N.A., Ruqia, B., Zahid, M.: Resource allocation using Fog-2-Cloud based environment for smart buildings. In: 14th IEEE International Wireless Communications and Mobile Computing Conference (IWCMC-2018) (2018)Google Scholar
  16. 16.
    Chekired, D.A., Khoukhi, L., Mouftah, H.T.: Decentralized Cloud-SDN architecture in smart grid: a dynamic pricing model. IEEE Trans. Ind. Inf. (2017)Google Scholar
  17. 17.
    Moghaddam, M.H.Y., Leon-Garcia, A., Moghaddassian, M.L.: On the performance of distributed and cloudbased demand response in smart grid. IEEE Trans. Smart Grid (2017)Google Scholar
  18. 18.
    Yi, G., Kim, H.-W., Park, J.H., Jeong, Y.-S.: Job allocation mechanism for battery consumption minimization of cyber-physical-social big data processing based on mobile cloud computing. IEEE Access 6, 21769–21777 (2018). ISSN 2169-3536Google Scholar
  19. 19.
    Devi, D.C., Uthariaraj, VR.: Load Balancing In Cloud Computing Environment Using Improved Weighted Round Robin Algorithm For Nonpreemptive Dependent TasksGoogle Scholar
  20. 20.
    Kumrai, T., Ota, K., Dong, M., Kishigami, J., Sung, D.K.: Multiobjective Optimization in Cloud Brokering Systems for Connected Internet of ThingsGoogle Scholar

Copyright information

© Springer Nature Switzerland AG 2019

Authors and Affiliations

  1. 1.COMSATS UniversityIslamabadPakistan
  2. 2.Government College UniversityLahorePakistan

Personalised recommendations