Efficient Load Balancing in Cloud Computing Virtual Machines via Dynamic Allocation, Research Paper Example

Abstract

This paper provides a method for dynamically assigning cloud resources. The required resources may be set up and booted by the user, and they have to pay for them because cloud computing will need to offer an efficient mechanism for managing and allocating resources in the future, Infrastructure as a Service’s dynamic scheduling and consolidation mechanisms. Instances may be created or removed on-the-fly depending on demand, and user needs using this technique. In addition, I would want to improve numerous aspects of cloud performance, such as throughput and virtual topologies depending on application requirements.

Introduction

Cloud computing makes computer and storage resources accessible to various consumers. The cloud uses the internet and distant servers to manage data and apps. Companies and individuals may access data and applications without installing software through any internet-enabled device may access data and files [1]. Cloud computing centralizes memory, computation, storage, and bandwidth. For example, Infosys produces cloud-based software leveraging Microsoft’s Windows Azure Cloud services, including SQL Data Services, to help auto dealers share inventory and other resources. This paper aims to assess the efficiency of load balancing in cloud computing virtual Machines via dynamic allocation.

Giftag makes use of Google App Engine to build and distribute desire lists culled from different sources [2]. Wang Fu Jing Department Store manages its supply chain using IBM Cloud services. Cloud computing services are classed as public, private, or hybrid. Rather, more advanced models abstract from simpler ones such as PaaS from IaaS. Software and product development contained tools on the provider’s structure.

People using PaaS may use APIs, internet portals, or client-side-made software to manage and operate such outlets [4]. The SaaS model The cloud approach connects users to their service provider through a front-end portal that supplies hardware and software. Infrastructure-as-a- Start and stop virtual server processes. Cloud computing allows enterprises to pay only for their capacity and add more. This pay-as-you-go concept is called utility computing. A lot of people are worried about Cloud Computing’s performance.

This study focuses on performance. Cloud computing does not solve application performance difficulties. In other words, the cloud may be slower than conventional on-premises solutions [6]. As a result, the cloud continually monitors all apps. Monitoring guarantees SLAs are satisfied while maximizing performance and uptime; furthermore, virtualization can efficiently handle dynamic resources in Cloud computing. For example, a virtual machine image may be produced and mapped to a real server for more demanding clients. This assists in resource diversity and platform insignificance.

One may dynamically remap VMs and real resources to balance system load using virtualization. These benefits make virtualization popular in Cloud computing. Install and utilize an operating system or application in a computer environment [7]. A virtualization layer that manages CPU, memory, storage, network, and other hardware resources is vital for the host OS protections. The virtualization layer may produce VM environments. Dynamic virtual machine allocation for load balancing.

Load Balancing

In order to ensure that no machine is idle, overburdened, or under-loaded, balancing is used. Load balancing helps the cloud reduce latency, increase throughput, and improve system stability [11]. On the other hand, scheduling tasks in this optimization technique is NP-hard; thus, job planning, resource allocation, and resource management are all part of these approaches. The load balancing survey papers could not classify all approaches and techniques. Problems with load balancing in the cloud are similar to those encountered in traditional architectures. Load unbalancing is discovered, along with suggestions for further development [9]. Analyze present-day solutions for load balancing and its use in cloud computing environments. Different load balancing methods and algorithms may be categorized. When designing a load-balancing algorithm, I have considered the difficulties academics face.

VM Load Balancing

OS and apps are isolated from the hardware using a virtual machine (VM). As part of the simulator, Cloud services represent different data centers, such as datacenter host components required to handle VM-sandboxed application elements [9]. The DataCenter object’s job is to manage the data center’s virtual machines (VMs) and redirect user requests to the VMs. The Data Center Controller uses a VM Load Balancer to deliver recommendations to VMs. The majority of VM load balancing solutions and monitored constantly. Datacenter controllers give an ID to create a Virtual Machine when a request comes in from a load balance throttled. Active Monitoring Load Balancer tracks the number of requests made to each VM. This algorithm discovers the least-used virtual machine (VM) when one is requested [5]. There can be only one winner if there is more than one. Throttling and active monitoring are two of the algorithms in use. Using a new method based on active monitoring load balancing can minimize response period, processing span, and overall cost.

VM Load Balancing Method

Job response time and utilization of resources have developed as a result of redistribution. In dynamic load balancing, the system’s present behavior is the only thing taken into account; the prior state is ignored [12]. Methods for estimating and comparing virtual machine workloads, as well as for transferring tasks across machines, should be included. Network load is measured in terms of delay. It takes time to complete a task in a single step. As a result, the project has several manageable chunks that influence its operations [6]. Technique for balancing the demands of a network to adapt to changing client demands while reducing resource use uses load balancing. Cloud computing efficiency and task response time are prioritized by first assigning workloads to virtual machines.

Recognize the CPU and RAM requirements for each VM and the CPU and RAM that is currently available. In the second step, the instance is either added or removed based on comparing the available resources to the requested resources. Cloud Watch is an engine for storing time-series data [4]. The monitoring store regularly receives data from servers and other services, which clients may query. The greatest way to organize is from the top down. Load-balancing strategies are difficult to classify since there is no specified hierarchical taxonomic classification. Algorithm nature, state, trait, load balancing type, and method are all categorized similarly. In contrast to earlier research, this one looks at the LB algorithms in great depth. Load balancing algorithms that are pre-emptive are good because it offers a broad taxonomic classification of data [2]. Algorithms for static, dynamic, and hybrid LBs Load balancing occurs through scheduling or allocation strategies. As a result, many different load balancing technologies are necessary.

Structure of The Algorithm

The classification of load balancing strategies was based on algorithm type. This classification separates LB algorithms into two groups: those that are proactive and those that are reactive [4]. However, communication routing protocols have undergone extensive study in other technology areas, such as mobile ad hoc networks (MANETs). Instead of just reacting to change, algorithms designed with a proactive LB perspective investigate how they may create it. Instead of waiting for an issue to occur, it aims to prevent it from happening. In order to be proactive, one must look for and take advantage of possibilities and anticipate potential problems and risks. Existing procedures use many aggressive methods and basic strategies, and there are no new ideas [7]. Some scholars propose an LB technique based on game theory that is fair to all participants playing in the game’s optimal load balancing Nash equilibrium.

Detailed Design

In the event of DNS or availability zone problems, the load balancing service distributes the load across many instances. Monitor AWS resources and client apps using Amazon Cloud Watch. Developers and system administrators may collect and manage metrics to keep applications and businesses running correctly [8]. Amazon CloudWatch may also monitor bespoke metrics produced by customers’ applications and services. With Amazon CloudWatch, you can monitor everything from resource use to application performance to operational health.

Experimental Setup

This section describes how to run workflows in an experimental environment. The VM load balancing method is implemented in Java. The most popular, feature-rich, and dependable applications in commercial Cloud Workflows are loosely connected parallel programs made out of data and control flow connections [12]. Tasks may also interact with one another through a network. In order to accommodate these differences, the experiments in this study use (a) single nodes for all testing and (b) EC2’s local disk for all tests. Although single-node trials do not assess the scalability of cloud services, they give an application-oriented understanding of the underlying resource capabilities [11]. Static, dynamic, and hybrid LB methods use different state information forms. LB is the approach to classifying using the LB algorithm that is most frequently used. This is where the majority of load balancing comparison studies begin. Static LB offers VM jobs to perform throughout the construction process. Despite their simplicity, static algorithms are challenging to balance [6]. To use static algorithms, one must be familiar with machines. Due to its inability to migrate during task execution is incompatible with distributed systems like the cloud.

Question Identification

A literature review highlighted a series of questions addressed before implementing load balancing. In literature, the following are the research questions. In addition, a collection of questions was created to highlight the necessity of load balancing in cloud computing.

• Q1: What creates load imbalance? This inquiry asks why load unbalancing occurs. Identification of variables causing load unbalancing. The technique of load balancing is inadequate without awareness of unbalancing factors.
• Q2: Why is load balancing critical in cloud computing? This research question aims to address the concerns of people offering cloud computing.
• Q3: How long does the load balancing algorithm take place. This question asks how long the load balancing algorithm takes to finish. The present research does not use algorithm complexity to categorize LB algorithms. However, to be valid, the algorithm must run at real-time complexity.

Research Methodology

The right research methods were needed to identify the fundamental causes of load balancing issues. Load balancing is based on algorithms, paradigms, theories, and approaches used to handle the literature study [11]. The GCF technique examines the unbalanced load issue by breaking it into its constituent parts, variables, and parameters. The literature review focuses on research relevant to cloud load balancing solutions, using Kitchenham’s study criteria for Systematic Literature Review (SLR). Other researchers may use an SLR to discover new information [9]. However, there are various reasons why IaaS cloud load balancing fails, thus why SaaS cloud load balancing is an issue.

Scheduling is NP-hard because of the unpredictable criteria of managing tasks and cloud traffic. However, no strong, precise, and effective mapper and generator function enhance scheduling. Load unbalancing happens when there is an uneven allocation of work among computer resources and their relationships [12]. All of these factors must be considered when selecting a load balancing strategy. Minimize wastage of resources and costs associated with migration and ensure service level agreement (SLAs). The loss of these characteristics leads to a decrease in CSC quality of service and a reduction in CSP profitability. CSPs have a problem balancing quality of service and service level agreements SLAs [1]. Therefore, it is necessary for researchers to concurrently increase performance and economic metrics to address NP-hard problems, such as load balancing. In light of these constraints, load balancing is a complicated issue to solve. By classifying load balancing approaches as either single- or multi-objective, this section aims to address.

It is impossible to find a load balancing strategy that considers all relevant parameters. Some advocated a single-target technique, while others tried to enhance many metrics. It may be challenging to deploy single-target load balancing solutions because of the complexity of the architectural design. Multi-objective approaches start with simple techniques [3]. The load balancing approach’s temporal complexity measures and should be used as a performance benchmark. Algorithmic complexity is only addressed in seven of the top 35 articles in this study, accounting for barely 20% of the whole search area.

It has been shown that dynamic algorithms may be classed as either offline (batch mode) or continuous online mode [4]. User jobs in batch mode are assigned a VM after they reach the scheduler, but in online mode, they are assigned a VM instantly. It is more difficult to handle incoming traffic with dynamic load balancing algorithms than static ones since the state of an ongoing operation might change. As the system’s present state changes, dynamic load balancing adjusts to meet the resulting increase in processing demand [5]. While dynamic load balancing enables work to be moved from one overburdened computer to another, it is substantially more complex to implement than static LB techniques since it is dynamic.

Data Collection and Search Processes

ACM Digital Library, Springer, and Elsevier are some of the databases used to gather information. I was able to find the information I needed thanks to a systematic search [11]. A lot of the advanced search query processing concepts started: in the actual processing of search queries themselves. By employing Boolean operations such as “OR” and “AND,” the search area for relevant data is reduced. Some examples include resource utilization, task migration, and task scheduling. Searches for reliable articles were conducted using regulated and uncontrolled keywords and a complicated filtering process [10]. Metadata alone and complete text with metadata are the two sub-options under the advanced keyword, or phrase option is better.

Results

Cloudsim, a cloud simulator, was used to test the results. The results show that the dynamic allocation works better than I had predicted. According to the analysis, migration time is reduced by pre-allocating virtual machines. A greener approach to cloud computing is provided by lowering the population of active hosts in the cloud using this method that has been proven through thorough testing. The analysis also indicates that relocation and idle time negatively impact this method. Checking whether the host is acceptable for VMs avoids migration due to excessive host demand. Another benefit of using this method to create a new virtual machine is that it does not strain the host. Rather than allow the host to overflow, a new Virtual Machine is created. I preferred the least-loaded host from the optimal host since data shows that less time is spent and more energy is saved through dynamic allocations.

Conclusion

Comparing load balancing methods in the literature shows various reasons CGF and SLR are used. The use of search engines and sophisticated filtering methods was beneficial [2]. According to the findings of this research, a multilayer taxonomy may be proposed based on five criteria. Statistically, proactive tactics are more dynamic than reactive ones, which are less active. Proactive strategies that are not dynamic are nevertheless proactive strategies [7]. The work schedule contributed 46 and 61.85 percent in proactive and reactive strategies, respectively.

I propose enhancements to the load balancing process as part of our investigation. Measures critical to the quality of service (QoS) are often overlooked. Peer-reviewed studies do not cover all of the most important Quality of Service indicators. For example, 70% of studies do not incorporate the algorithm’s complexity when evaluating the performance of a load balancing solution [6]. There is a wide range of current load balancing methods that employ simulators. With this in mind, it may be possible to address future challenges in load balancing by developing an efficient and sophisticated algorithm that considers new QoS metrics and algorithm complexity assessment [10]. The taxonomy provides numerous ways to assist future researchers in dealing with unbalanced loads, especially algorithms drawn from nature, artificial intelligence (AI), or arithmetic.

Works Cited

[1] S. Mishra, B. Sahoo, and P. Parida, “Load balancing in cloud computing: A big picture,” Journal of King Saud University – Computer and Information Sciences, vol. 32, no. 2, pp. 149-158, 2020. Available: 10.1016/j.jksuci.2018.01.003 [Accessed 22 March 2022].

[2] R. P. and V. M., “An Efficient Dynamic Load Balancing Mechanism for Cloud Computing Environment,” 2021. Available: 10.21203/rs.3.rs-230430/v1 [Accessed 22 March 2022].

[3] M. Adhikari and T. Amgoth, “Heuristic-based load-balancing algorithm for IaaS cloud,” Future Generation Computer Systems, vol. 81, pp. 156-165, 2018. Available: 10.1016/j.future.2017.10.035 [Accessed 22 March 2022].

[4] M. Kumar, K. Dubey, and S. Sharma, “Elastic and flexible deadline constraint load Balancing algorithm for Cloud Computing,” Procedia Computer Science, vol. 125, pp. 717-724, 2018. Available: 10.1016/j.procs.2017.12.092 [Accessed 22 March 2022].

[5] S. Chen, Y. Chen and S. Kuo, “CLB: A novel load balancing architecture and algorithm for cloud services,” Computers & Electrical Engineering, vol. 58, pp. 154-160, 2017. Available: 10.1016/j.compeleceng.2016.01.029 [Accessed 22 March 2022].

[6] M. Kumar and S. Sharma, “Dynamic load balancing algorithm for balancing the workload among virtual machine in cloud computing,” Procedia Computer Science, vol. 115, pp. 322-329, 2017. Available: 10.1016/j.procs.2017.09.141 [Accessed 22 March 2022].

[7] L. Tang et al., “Online and offline based load balance algorithm in cloud computing,” Knowledge-Based Systems, vol. 138, pp. 91-104, 2017. Available: 10.1016/j.knosys.2017.09.040 [Accessed 22 March 2022].

[8] A. Singh and S. Prakash, “WAMLB: Weighted Active Monitoring Load Balancing in Cloud Computing,” Advances in Intelligent Systems and Computing, pp. 677-685, 2017. Available: 10.1007/978-981-10-6620-7_65 [Accessed 22 March 2022].

[9] Z. Xiao, Z. Tong, K. Li and K. Li, “Learning non-cooperative game for load balancing under self-interested distributed environment,” Applied Soft Computing, vol. 52, pp. 376-386, 2017. Available: 10.1016/j.asoc.2016.10.028 [Accessed 22 March 2022].

[10] N. Antonopoulos and L. Gillam, Cloud computing. Cham: Springer, 2017.

[11] W. O’Brien and S. Ponticelli, Computing in civil engineering 2015.

[12] E. Jafarnejad Ghomi, A. Masoud Rahmani and N. Nasih Qader, “Load-balancing algorithms in cloud computing: A survey,” Journal of Network and Computer Applications, vol. 88, pp. 50-71, 2017. Available: 10.1016/j.jnca.2017.04.007 [Accessed 22 March 2022].