Energy-aware Load Balancing Techniques for Cloud Computing

Abstract

Cloud computing, a form of distributed computing, promises to deliver reliable services through next-generation data centers, built on virtualized compute and storage technologies. Cloud providers rely on large data centers to o er resources required by users but the energy consumed by cloud infrastructures has lately become a key environment and economic concern. Lot of energy is wasted in these data centers due to the under-utilized resources and therefore these under-utilized resources should be e ciently utilized to conserve energy. Another important concern in cloud computing is load balancing. Load Balancing is essential for distributing dynamic workloads over multiple nodes so that it is ensured that no single node is overwhelmed. It thus helps in avoiding hot-spots and enhancing resource utility levels thereby, reducing energy consumption. As energy e ciency has appeared as the utmost essential design requirement for current computing systems, it determines the need of energy-aware load balancing in clouds to overcome the issue of energy-e ciency. This research work proposes two Energy-aware Load Balancing (ELB) techniques. These techniques are based on the proposed Energy-aware Resource Utilization (ERU) model and the Fire y Optimization based Energy-aware Virtual Machine Migration (FFO-EVMM) technique. The proposed ERU model e ciently manages cloud resources and enhances their utilization by allocating the jobs to the appropriate resources, using the Arti cial Bee Colony (ABC) optimization approach. It also maximizes the energy-e ciency of the cloud data centers through optimal resource usage, without degrading the system performance. Thereafter, the FFO-EVMM technique has been proposed that migrates the maximally-loaded VM to the least energy-consuming node, to reduce the consumed energy in the cloud data centers at run-time. The proposed technique intends to enhance the energy-e ciency through optimum migration of VMs, thereby improving resource utilization levels. Finally, the ELB model has been proposed that helps the proposed load balancing techniques to take energy-aware decisions, from the perspectives of, both the provider and the user. The two ELB techniques, namely ELB(RU) and ELB(w/o RU), aid the system administrator to balance system's load, across the minimum required active nodes, in an energy-aware manner. The ELB(RU) technique uses an amalgamation of ERU & FFO-EVMM. Through ABC-based ERU, the ELB(RU) technique aspires to maximize resource utilization by appropriately assigning the users workloads to the energy-aware nodes. Furthermore, by using the FFO-EVMM technique, it migrates the most loaded VMs to the least energy-consuming nodes, thereby curtailing the energy needs of the cloud data centers. The ELB(w/o RU) technique strives to augment the performance of the system by processing maximum number of workloads. It uses only the FFO-EVMM algorithm for energy-aware VM migrations, consequently avoiding hot-spots and bringing down the energy consumption levels in the cloud data centers. The performance of the proposed ABC-based ERU model and the FFO-EVMM technique has been evaluated by comparing them with the First Fit Decreasing (FFD) and the Ant Colony Optimization (ACO) algorithms, through CloudSim toolkit. The results demonstrate that an average of 8.68% of nodes and 8.66% of energy have been conserved using ERU over FFD and ACO-based techniques. Whereas, an enhancement in the average energy consumption of about 44.39% has been attained by reducing an average of 72.34% of migrations and saving 34.36% of hosts, thereby, making the data center more energy-aware. A case study has been conducted at the data center of a telecom service provider \Bharat Sanchar Nigam Limited (BSNL), Chandigarh, India", to test and validate the proposed ELB techniques. The competence of the proposed ELB(RU) technique is exhibited by comparing it with the Round Robin (RR) technique which is currently being used in BSNL data center and also by comparing it to FFD & ACO. The adequacy of the proposed ELB(RU) technique is accomplished by saving 40.47% of the average energy consumption, enhancing 49.68% of CPU utilization level, 24.41% of memory utilization level, by lowering 63.10% of VM migrations and reducing 53.21% of nodes. Whereas, the e ectiveness of the ELB(w/o RU) technique is presented by comparing it with RR, FFD, ACO & ELB(RU). The results demonstrate that ERU(w/o RU) enhances the average performance by 38.55%. It further attains 20.53% of drop in the average energy consumption by reducing 45.57% of average VM migrations and by saving 19.64% of nodes on an average. The improved results illustrate that the proposed ELB techniques e ectively balance the load, enhance resource utilization and performance levels, thereby curtailing the levels of energy consumption in the cloud data centers. The empirical evaluation clearly justi es the role of ELB(RU) to be more useful for aggrandizing the resource utilization levels and ELB(w/o RU) to be more e ective in augmenting the performance of the system. As per the Cloud user or Cloud provider requirement, the appropriate algorithm can be chosen. The obtained results have been veri ed statistically by using the Coe cient of Variance (CoV), to approve the stability of the proposed ELB techniques. By reducing consumed energy, all the proposed techniques, indirectly reduce carbon emissions and cooling requirements of the cloud data centers, leading to a further drop in the energy demand and helping in achieving green computing.