BACOS: A Dynamic Load Balancing Strategy for Cloud Object Storage

• Manoel Rui P. Paula
• Eduardo Rodrigues
• Victor A. E. Farias
• Flávio R. C. Sousa
• Javam C. Machado

Resumo

Cloud Computing is an efficient model for processing and storing large amounts of data. The cloud is composed by heterogeneous resources and has a variable workload. Cloud object storage systems arise as an efficient manager for data using heterogeneous devices, regarding storage capacity and performance. In the cloud, since workload changes dynamically, the dynamic reconfiguration is needed to improve resource utilization. Thus, load balancing techniques are crucial to redistribute workload among the processing nodes to avoid underloading or overloading. Conventional load balancing strategies are only aware of storage devices' capacity, resulting in system performance degradation. To address these limitations, this paper presents a non-intrusive approach to load balancing in the cloud which considers storage devices with heterogeneous performance. Experimental results confirm that our approach improves performance in terms of response time and throughput when compared to the strategy employed by Openstack Swift object storage.