The advent of cloud computing involving virtualization technologies has offered maximum opportunities for hosting low-cost virtual resources without any infrastructure. The cloud data centers generally consist of heterogeneous commodity servers that are capable of hosting multiple Virtual Machines (VMs) with significantly varying specifications and dynamic resource utilization potentialities. In this context, servers hosting heterogeneous VMs with potentially varying specifications cannot handle unpredictable and variable workloads leading to an imbalance in resource utilization on the server causing Service Level Agreement (SLA) violations and degradation in performance. The cloud data centers are highly unpredictable and dynamic due to the fluctuating resource utilization of VMs, irregular resource utilization patterns of cloud consumers constantly requesting VMs, great deviation in the hosts’ performance in the process of handling different levels of load and unstable arrival and departure rate of data center consumers. These situations are responsible for introducing unbalanced loads in the data center of the cloud that results in SLA violations and performance degradation. Moreover, this imbalanced resource utilization is seen in most of the cases when a VM executes computation-rich applications in spite of its low memory requirements. This problem of resource utilization has proved to be a non-deterministic polynomial time hard problem which can be predominantly solved by hybrid metaheuristic approaches. In this paper, an Improved Artificial Bee Colony using Monarchy Butterfly Optimization Algorithm-based Load Balancıng (IABC-MBOA-LB) is proposed for effective resource utilization in clouds. The proposed IABC-MBOA-LB includes global exploration capability of ABC and local exploitation potential of MBOA for effective allocation of user tasks to VMs. It focuses on network and computing resources in order to prevent fragmentation and unnecessary increase in the task finishing times as both should be potentially explored for better resource allocation process. The simulation experiments of the proposed IABC-MBOA-LB scheme confirm its predominance in minimizing load variance and standard deviation of utilization, makespan, standard deviation of connections, average imbalance degree and maximizing throughput independent of the number of tasks and VMs in the cloud.

涉及虚拟化技术的云计算的出现为在没有任何基础架构的情况下托管低成本虚拟资源提供了最大的机会。云数据中心通常由异构商品服务器组成，这些服务器能够托管规格和动态资源利用潜力明显不同的多个虚拟机（VM）。在这种情况下，承载具有可能变化的规格的异构VM的服务器无法处理不可预测的可变工作负载，从而导致服务器上资源利用率的不平衡，从而导致违反服务水平协议（SLA）和性能下降。由于VM的资源利用率不断变化，因此云数据中心具有高度的不可预测性和动态性，云用户不断请求虚拟机的不规则资源利用模式，在处理不同级别的负载的过程中主机性能存在很大偏差，数据中心用户的到达和离开速率不稳定。这些情况负责在云的数据中心中引入不平衡的负载，从而导致违反SLA和性能下降。此外，尽管VM的内存需求较低，但在大多数情况下，当VM执行富含计算的应用程序时，就会看到这种资源利用率不平衡的情况。这个资源利用问题已被证明是一个不确定的多项式时间难题，主要可以通过混合元启发式方法解决。在本文中，为了有效利用云资源，提出了一种基于君主蝴蝶优化算法的基于负载均衡的改进人工蜂群（IABC-MBOA-LB）。拟议中的IABC-MBOA-LB包括ABC的全球勘探能力和MBOA的本地开发潜力，可将用户任务有效地分配给VM。它着重于网络和计算资源，以防止碎片化和不必要的任务完成时间增加，因为应同时探索两者，以实现更好的资源分配过程。所提出的IABC-MBOA-LB方案的仿真实验证实了其在最小化负载方差和利用率的标准偏差，建立时间，连接的标准偏差，平均不平衡程度以及最大化吞吐量方面的优势，而不受云中任务和VM数量的影响。