BACOS: uma estratégia de balanceamento de carga para sistemas de armazenamento de objetos em nuvem

Abstract

Cloud computing is an emerging and efficient computing model for processing and storing large amounts of data. Cloud storage service providers use heterogeneous storage devices as a way to extend the resources of a storage system by considering the best tradeoff between maintenance and performance costs. Cloud object storage systems emerge as scalable solutions and efficient data managers using heterogeneous devices, in terms of storage capacity and performance. In the cloud, as the workload changes, dynamic matching between load and storage device capabilities is needed to improve resource utilization and optimize the overall performance of an object storage system. Thus, load balancing techniques are crucial for redistributing the workload among processing and storage nodes to avoid underloading or overloading. Most conventional load balancing strategies in cloud storage systems are only aware of the storage capacity of storage devices or make assumptions about them being homogeneous, resulting in degradation of the storage system. To address these limitations, this work presents a non-intrusive load balancing strategy called BACOS that takes advantage of storage devices with heterogeneous performance in a cloud storage system. The results of the experimental evaluation confirm that BACOS can improve the performance of an object storage system in terms of response time, throughput and success rate of read/write requests in scenarios that vary the workload.