Dynamic Load Balancing (DLB) is a technique implied in cloud computing to distribute workloads dynamically across available resources to guarantee optimal performance, reliability and scalability. Dynamic load balancing works correspondent to current system behavior and workload. Dynamic load balancing can handle almost large fluctuations in workload and system behavior. It involves continuously monitoring the utilization of resources and automatically reallocating tasks to underutilized resources. DLB techniques can be broadly categorized into two categories:
• Client-side - Client-side DLB involves distributing tasks among resources at the client end
• Server-side - Server-side DLB includes distributing tasks among resources at the server end.
There are two approaches to dynamic load balancing :
• Distributed :
A distributed approach spreads the load-balancing overhead over numerous computing nodes. So there is no single point of failure risk as there are multiple participating nodes.
• Non-distributed :
A centralized load balancing approach is directed by a central control unit and is effortless to implement but carries the hazard of a single point of failure. Further distributed centralized dynamic load balancing is categorized into two divisions:
• Centralized:
A single node is designed and answerable for the entire system.
• Semi-centralized:
In semi-centralized, the nodes are divided into clusters, and the individual cluster has its load balancer. A cluster is considered the master cluster, and the particular node is responsible for the entire system.
• Round Robin: A simple technique of assigning the task to available resources in a cyclic order.
• Least Connections: Tasks with the least active connections are assigned to the resource.
• Weighted Round Robin: Similar to Round Robin, resources are assigned different weights for the capacity.
• IP Hash: Tasks are assigned to resources based on a client-s IP address hash.
• CPU utilization: Tasks are assigned to the resource with the lowest CPU utilization.
• Adaptive Load Balancing: An advanced technique that uses machine learning algorithms to dynamically adjust the load balancing strategy based on resource utilization, network traffic and other factors.
• Enhanced User Experience: Improving performance and reducing downtime, DLB enhances the user experience for cloud-based applications.
• Minimize construction time
• Limited Customization: DLB techniques may not provide the customization required for specific applications, as they are designed to work with a broad range of workloads.
• Limited Adaptability: DLB techniques may not be able to adapt to changes in resource utilization patterns or workflows, requiring manual intervention to adjust the load-balancing strategy.
• Inadequate Resource Management: DLB techniques may not provide adequate resource management capabilities, as they are designed primarily to distribute workloads.
• Balancing Local and Global Optimization: DLB must balance local and global optimization, as optimizing resource utilization at the local level may not result in optimal performance at the global level.
• Managing Resource Constraints: DLB must manage constraints on resources, such as available memory, CPU cycles, and network bandwidth, to ensure that workloads are executed efficiently.
• Ensuring Security: DLB must ensure the security of workloads and resources as tasks are moved between resources in a cloud environment.
• Ensuring Fairness: DLB must ensure that workloads are distributed fairly among resources, which can be difficult to achieve in a cloud environment with varying resource capacities.
• Context-aware DLB: Context-aware DLB techniques consider contextual information, such as the location of resources and users, to optimize resource allocation and improve performance.
• Predictive DLB: Predictive DLB techniques use historical data to predict future resource utilization patterns and allocate resources accordingly, reducing latency and improving performance.
• Autonomic DLB: Autonomic DLB techniques aim to make DLB systems self-governing, reducing the need for manual intervention and improving reliability.
• Machine Learning-based DLB: Machine learning algorithms can dynamically adapt DLB techniques based on resource utilization patterns, workloads, and other factors, providing more efficient and effective resource allocation.
• Integration with Resource Management: Research focused on integrating DLB with resource management systems, including using cloud orchestration and management platforms.
• Energy-efficient DLB: Research focused on developing energy-efficient DLB techniques that reduce energy consumption while maintaining performance.
• Performance Evaluation: Research focused on evaluating DLB techniques in cloud computing environments, including using simulation and modeling tools to compare different DLB strategies.
