Comparative Analysis of Existing Dynamic Load
Balancing Techniques
Nayandeep Sran
CSE department PEC University of Technology
Chandigarh
Navdeep Kaur
Assistant Professor, IT department PEC University of Technology
Chandigarh
ABSTRACT
The anticipated uptake of Cloud computing, built on well-established research in Web Services, networks, utility computing, distributed computing and virtualization, will bring many advantages in cost, flexibility and availability for service users. Cloud is based on the data centers which are powerful to handle large number of users. As the cloud computing is a new style of computing over internet, it has many advantages along with some crucial issues to be resolved in order to improve reliability of cloud environment. Central to this is the implementation of an effective load balancing algorithm. This paper investigates two distributed load balancing algorithms which have been proposed for load balancing: round robin and throttled scheduling.
General Terms
Cloud computing, load balancing.
Keywords
Virtual machine, round robin scheduling, throttled scheduling, simulation, User Base(UB),Data Center(DC).
1.
INTRODUCTION
Cloud computing mean different things to different people. Cloud computing is an internet-based model of computing, where the shared information, software and resources are provided to computers and other devices upon demand. The prominent attributes are on-demand scalability of highly available and reliable pooled computing resources, secure access to metered services from nearly anywhere, and dislocation of data from inside to outside the organization. While aspects of these characteristics have been realized to a certain extent, cloud computing remains a work in progress [1]. Cloud computing is a service oriented architecture [3], which is provided via internet. Many companies are trying to implement and introduce clouds, due to its simple and flexible architecture. Although clouds are bifurcated in public private and hybrid models but still problem of reliability may arise in these. Cloud models used virtualization technology; this technology helps in slicing a single data centre or high power
server to act as multiple machines. It depends on the hardware configuration of the data centre or server in how may virtual machine they can be divided [4].
As the cloud computing is a new style of computing over internet. It has many advantages along with some crucial issues to be resolved in order to improve reliability of cloud environment. These issues [5] are related with:
Security
Efficient Load Balancing Performance Monitoring
Consistent & Robust Service abstractions Resource Scheduling
Scale and QoS management Interoperability & Portability
Requires a constant & speedy Internet connection
[image:1.595.334.540.367.517.2]Central to these issues lays the establishment of an effective load balancing algorithm. The load can be CPU load, memory capacity, and delay or network load [2].
Figure 1 shows classification of load balancing algorithms. Load balancing is the phenomenon of distributing the load among various nodes of a distributed system so as to improve both resource utilization and job response time whilst also avoiding a situation where some of the nodes are heavily loaded while other nodes are not doing any work [8]. Load balancing guarantees that all the processor in the system or every node in the network does approximately the equal amount of work at any instant of time [6]. Load balancing algorithms are categorized as: The important things to consider while developing such algorithm are : estimation of load, comparison of load, stability of different system, performance of system, interaction between the nodes, nature of work to be transferred, selecting of nodes and many other ones. This load considered can be in terms of CPU load, amount of memory used, delay or Network load [7]
26 The motivation of our research is to consider the
distributed cooperative load balancing algorithm in which all nodes present in the system execute the algorithm and the task of load balancing is shared among them such that the nodes work side-by-side to achieve a common objective to improve the overall response time, processing time etc. Benefit of dynamic load balancing is that if any node is fail it will not halt the whole network [8]
2.
LITERATURE REVIEW
M. Randles et al. [10] investigated a decentralized honeybee-based load balancing technique that is a nature-inspired algorithm for self-organization. It achieves global load balancing through local server actions. Performance of the system is enhanced with increased system diversity but throughput is not increased with an increase in system size. It is best suited for the conditions where the diverse population of service types is required [9]
M. Randles et al. [10] investigated a self-aggregation load balancing technique that is a self-aggregation algorithm to optimize job assignments by connecting similar services using local re-wiring[9] The performance of the system is enhanced with high resources thereby in-creasing the throughput by using these resources effectively. It is degraded with an increase in system diversity.
Y. Lua et al. [11] proposed a Join-Idle-Queue load balancing algorithm for dynamically scalable web services. This algorithm provides large-scale load balancing with distributed data centers by, first load balancing idle processors across dispatchers for the availability of idle processors at each dispatcher and then, assigning jobs to processors to reduce average queue length at each processor. By removing the load balancing work from the critical path of request processing, it effectively reduces the system load, incurs no communication overhead at job arrivals and does not increase actual response time [9].
Authors in [12] have recommended load balancing in a three–level cloud computing network, by using a scheduling algorithm which combines the features of Opportunistic Load Balancing (OLB) and Load Balance Min-Min (LBMM) which can utilize better executing efficiency and maintain load balancing of the system. The objective is to select a node based for executing the complicated tasks that needs large-scale computation. The scheduling algorithm proposed in this paper is not dynamic and also there is an overhead involved in the selection of the node.
CloudSim enables seamless modeling, simulation, and experimenting on Cloud computing infrastructures, which is based on java programming language. It allows us to study various constraints by changing the parameters. It provides a virtualization engine with extensive features for modeling life-cycle management of virtual machines in a data center, including policies for provisioning of virtual machines to hosts, scheduling of resources of hosts among virtual machines, scheduling of tasks in virtual machines, and modeling of costs incurring in such operations [13]. CloudSim allows simulation of scenarios modeling Infrastructure as a Service, Platform as a Service, and Software as a Service, because it offers basic components such as Hosts, Virtual Machines, and applications that model the three types of services [13].
Cloud Analyst is built directly on top of CloudSim toolkit, leveraging the features of the original framework and
extending some of the capabilities of CloudSim such as it enables a modeler to repeatedly execute simulations and to conduct a series of simulation experiments with slight parameters variations in a quick and easy manner[2][13]
3.
PROBLEM DEFINATION
The random arrival of load in heterogeneous environment can cause some server to be heavily loaded while other server is idle or only lightly loaded. Equally load distributing improves performance by transferring load from heavily loaded server. Efficient scheduling and resource allocation is a critical characteristic of cloud computing based on which the performance of the system is estimated. Load balancing is used to achieve a high user satisfaction and resource utilization ratio, making sure that no single node is overwhelmed, hence improving the overall performance of the system. The considered characteristics have an impact on cost optimization, which can be obtained by improved response time and processing time.
4.
DISTRIBUTED LOAD
BALANCING ALGORITHMS
All the above discussed algorithms in section 2 are not suitable for heterogeneous nature of cloud computing environment. So in this section we will discuss distributed workload of multiple network links to maximize throughput, minimize response time and to avoid overloading. We will be using three algorithms to distribute the load and will try to analyze the improvement in performance time and cost.
4.1
Round Robin Algorithm:
Round robinalgorithm is random sampling based. It means it selects the load randomly in case that some server is heavily loaded or some are lightly loaded. This algorithm simply allots the job in round robin fashion which doesn't consider the load on different machines [4]. In RR scheduling fairness is given to each process, i.e. processes get fair share of CPU because of given time slice [14]. It gives low turnaround time and average waiting time.
4.2
Throttled Load Balancing Algorithm: Throttledalgorithm is completely based on virtual machine. In this client first requesting the load balancer to check the right virtual machine which access that load easily and perform the operations which is given by the client or user. The job manager is having a list of all virtual machines, using this indexed list, it allot the desire job to the appropriate machine. If the job is well suited for a particular machine than that job is, assign to the appropriate machine. If no virtual machines are available to accept jobs then the job manager waits for the client request and takes the job in queue for fast processing.
5.
EXPERIMENTAL WORK
configure simulation, define Internet characteristics and run simulation [8], [9].This menu is for configuring the experiment and options to switch algorithm according to the requirement. Table 1 depicts the user base configuration along with its region boundaries.
Figure 2 shows the cloud analyst interface in which we have shown the running simulation showing connection between various user bases and data centers.
Table 1: Data center configuration
Data Centers along with region boundaries and region boundaries is also shown in the Figure 1. Table 2 shows our experimental setup .In this we will be using 5 data centers, 1000 user bases, 500 instructions which are executable per request and 50 virtual machines each for round robin and throttled load balancers, so as to differentiate between the results.
Figure 2: Cloud Analyst GUI [13]
6.
RESULTS
After performing experiment on the simulation we get the User Base (UB) which will communicate with the data centers to generate output. The average, minimum and maximum requesting time on the basis of overall response time is calculated and data center will generate desired outputs for the algorithms accordingly. The configuration assigned in terms of cost, data transfer is also shown in Table 2. And the results for cost calculation are shown in graph form in figure 3. costs are assigned in the dollar currency. Different user based processing time is also calculated using simulation. The data transfer cost after Appling the algorithm is also calculated and shown in table 2. And the comparison for cost is shown in figure 3.
7.
CONCLUSIONS
The request time and performance for the two policies applied (Round Robin and Throttled Load balancing) are shown. The cost analysis shown for the two algorithms is calculated in the experimental work in Section 5. The total cost calculated for Round Robin is more as compared to Throttled Load balancing algorithm, so Throttled Load balancing algorithm works more efficiently in terms of cost for load balancing on cloud data centers.
[image:3.595.328.560.71.288.2] [image:3.595.113.256.208.408.2]28
Table 2: Comparison of results round robin and throttled by using CloudSim
8.
FUTURE WORK
Cloud Computing is a perpetual research as all its aspects haven’t unfolded yet. At this time our paper is able to differentiate between the two load balancing algorithms which are round robin and throttled. In this paper we had done load balancing on the basis of performance and cost. Load balancing ensures that all the processors in the system in the network do approximately the similar amount of work at any instant of time. But here we find in this paper that even if throttled load balancer is providing better results, but still we can’t use throttled in large scale organizations because if number of requests increase beyond this threshold value in all available virtual machines then the requests are queued until a VM becomes available. In future we will try to find a new and advanced algorithm which will give better results on the basis of both cost and performance than both of the round robin and throttled load balancers.
Figure 3: Comparison of cost on the basis of response time
9.
REFERENCES
[1] Wayne Jansen, Timothy Grance, 2011 Guidelines on Security and Privacy in Public Cloud Computing, NIST Draft Special Publication 800-144.
[2] Jaspreet Kaur, 2012 Comparison of load balancing algorithms in a cloud, IJERA vol. 2.issue 3.
[3] Wei-Tek Tsai, Xin Sun, Janaka Balasooriya,2010 Service-Oriented Cloud Computing Architecture, Computer society
[4] Dr. Hemant S. Mahalle, Prof. Parag R. Kaveri and Dr.Vinay Chavan, 2013 Load Balancing in Cloud Data Centers.
[5] Ramgovind S, Eloff MM, Smith E, 2010 The management of security in cloud computing, IEEE.
[6] Ram Prasad Padhy, P Goutam Prasad Rao, 2011 Load balancing in cloud computing systems, National Institute of Technology, Rourkela
[7] Yatendra Sahu, R.K. Pateriya, 2013 Cloud Computing Overview with Load Balancing Techniques, IJCA Volume 65– No.24.
[8] Ajith Singh. N, M. Hemalatha, 2012 An Approach on Semi-Distributed Load Balancing Algorithm for Cloud Computing System IJCA Volume 56– No.12.
[9] Nidhi Jain Kansal, Inderveer Chana, 2012“Existing Load Balancing Techniques In Cloud Computing: A Systematic Re-View”, JISC Volume 3, Issue 1.
[10] Randles M., Lamb D. and Taleb-Bendiab A, 2010 24th Inter-national Conference on Advanced Information Networking and Applications Workshops, 551-556.
Parameters
Round Robin Load balancing
Throttled load balancing
Data centers 5 5
User base 1000 1000
Executable instruction length per request
500 500
Virtual machines 50 50
Avg(ms) Overall response time
101.04 120.44
Min(ms) Overall response time
49.13 49.13
Max(ms) Overall response time
129.80 241.96
Avg(ms) data center processing time
51.68 39.99
Min data center processing time
5.38 0.00
Max data center processing time
80.00 80.00
Total virtual machine cost($)
21.08 20.48
Total data transfer cost($) 0.29 0.29
[11] Lua Y., Xiea Q., Kliotb G., Gellerb A., Larusb J. R. and Greenber A, 2011 International Journal on Performance evaluation.
[12] Shu-Ching Wang, Kuo-Qin Yan, Wen-Pin Liao, Shun-Sheng Wang, 2010 Towards a Load Balancing in a Three-level Cloud Computing Network,” IEEE, pp. 108-113
[13] Bhathiya Wickremasinghe, Rodrigo N. Calheiros, and Rajkumar Buyya,CloudAnalyst: A
CloudSim-based Visual Modeller for Analyzing Cloud Computing Environments and Applications” from http://www.cloudbus.org/cloudsim/
